the structured credit handbook - wiki-bazarjosh/tsb/creditderivitive... · collateralized debt...

See the Disclosure Appendix for the Analyst Certification and Other Disclosures.

G L O B A L APRIL 28, 2004

S T R U C T U R E D C R E D I T

R E S E A R C H

Structured Credit Products GLOBAL

Arvind Rajan (212) 816-9904 [email protected]

New York

Glen McDermott (212) 816-8631 [email protected]

New York

Terry Benzschawel (212) 816-8071 [email protected]

New York

Dennis Adler Urvish Bidkar John Carpenter William E. Deitrick Alexei Kroujiline Pradeep Kumar David Li Robert Mandery Graham Murphy Ratul Roy Richard B. Salditt Jure Skarabot Glen Taksler Milena Todorova

This report can be accessed electronically via:

FI Direct

Yield Book

E-Mail

Please contact your salesperson to receive fixed-income research electronically.

The Structured Credit Handbook

Citigroup Global Markets 2

APRIL 28, 2004


April 28, 2004 The Structured Credit Handbook


Contents

Editor’s Note.......................................................................................................................................... 3

Collateralized Debt Obligations 7 The CLO Handbook 9 The CDO of ABS Handbook 47 Synthetic Arbitrage CDOs 81 The ABCs of CDO Equity 109 CDO Combination Securities 139 The CDO Trustee Report 163

Credit Derivatives 175 A Primer on Single-Tranche CDOs 177 Bull and Bear in a Boxx — Using Tranched Product to Express Credit Views 203 Trading the CDX.EM iBoxx 221 New Roll of the iBoxx CDX Indexes 231 Citigroup’s Primer on Credit Default Swaptions 239




Editor’s Note

In the pages that follow, we have gathered some of the most popular structured credit primers that we have written over the past four years. Much of the material contained in these pages is brand-new, including research on single-tranche CDOs, credit default swaptions and several sections on the new iBoxx indexes. Although some of the material is dated, many of the concepts, risks, structures, and analytical techniques described herein are as relevant today as the day when we first wrote about them. We hope you enjoy this broad-ranging structured credit products research compilation!

Arvind Rajan (212) 816-9904

Glen McDermott (212) 816-8631

Terry Benzschawel (212) 816-8071



Collateralized Debt Obligations

FEBRUARY 2004

Glen McDermott William E. Deitrick Alexei Kroujiline Robert Mandery

The CLO Handbook Customized Exposure to a Stable Asset Class



Su

mm

ary

The impressive growth that the leveraged loan market has displayed over the past decade has been accompanied by greatly improved liquidity and transparency. The benefits of this asset class, which include stable prices and high recovery and prepayments rates, can be accessed efficiently by CLOs. During the previous credit cycle, CLOs on average have demonstrated more stable performance than both high-yield bond CDOs and straight corporate debt. This stability has fueled CLO growth: CLOs now account for almost one-third of the primary CDO market.

The leveraged loan market has exhibited significant growth and expansion over the past decade, with $166 billion of new issuance in 2003 alone. The liquidity and transparency of the loan market continue to improve with the number of investors increasing from 14 to 372 over the past ten years. Despite this impressive growth, certain impediments remain.

Investors have been drawn into market by the advantageous characteristics of leveraged loans, which include floating interest rates, discount pricing, high prepayment rates, and greater control over the borrower in times of stress. The long-term, historical track record is strong: analysis shows that loans have performed solidly under various economic conditions, revealing a history of stable prices and robust recovery rates.

The stability of loans and inefficiencies of the loan market make leveraged loans particularly attractive to collateralized loan obligation (CLO) investors, who rely upon the asset-backed structuring technologies to gain leveraged exposure to this market. CLOs represent a subcategory of the collateralized debt obligation (CDO) market, which has grown by more than 1000% since 1995 to reach $70 billion of new issuance in 2003. CLOs now compose almost one-third of the primary CDO market

The primary reason for CLO market growth has been the strong performance of this asset class through the last credit cycle: loans lend themselves to leverage in the CLO context. Over the past seven years, on average, CLOs have exhibited higher ratings stability relative to both high-yield bond CDOs and straight corporate debt. Drivers of this stability include high recovery and prepayment rates, price stability, and CLO manager expertise.

Despite these obvious benefits, challenges remain. The recent surge in demand for institutional loans by structured vehicles has resulted in a significant tightening of the loan spreads with some CLOs containing an unacceptably high level of loans purchased at a premium. In addition, investors should be aware that a CLO portfolio built solely with broadly syndicated institutional loan tranches may contain significant name overlap. Recognizing this risk, CLO market participants are now looking to other sectors of the leveraged loan market in search of alternative collateral assets. Revolving credit obligations, middle-market loans, and European leveraged loans provide promising opportunities for further diversification, offering comparable yield and credit stability.



Leveraged Loan Market Overview

Strong Primary Market Growth Syndicated loans can be segmented into two market categories, leveraged (or high yield) and investment grade. The former is comparable to the high-yield bond market from a ratings and issuer leverage perspective. In most instances, a loan will be classified as a leveraged loan if it generally meets any of the following criteria: (1) debt ratings of below Baa3/BBB- from Moody’s and S&P, respectively; (2) debt/EBITDA ratio of 3.0 times or greater; or (3) at issue pricing of at least 125bp over LIBOR. Furthermore, leveraged loans are further classified as leveraged and highly leveraged.

Figure 1. Annual Leveraged Loan New Issue Volume, 1993–2003 (Dollars in Billions)

2847

6046

34

59

91

192 184

81 75

28

81101

135

220

256243

185

139 139

166

139

105

209

0

100

200

$300

1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003

Institutional

Pro RataTotal

Source: Standard & Poor’s.

Leveraged loans constitute a significant part of the syndicated loan market: new issuance was $166 billion in 2003, up from $139 billion in 2002 and greater than the 2003 high-yield bond market issuance of $137 billion. This loan issuance was composed of $91 billion institutional term loans and $75 billion pro rata loans (see Figure 1).

During the past decade, institutional investors have driven leveraged loan market growth and have made the fully drawn term loan (or “institutional” term loan) the most widely used structure in the leveraged loan market. New issue volume for the institutional term loan tranche set a record of $91 billion in 2003, or 55% of the total leveraged loan market. This level was up considerably from the $59 billion issued in 2002 and the prior record of $60 billion in 1999.

With respect to the overall market, the total amount of institutional leveraged loans outstanding is estimated at $148.1 billion, up $15.6 billion from 2002 (see Figure 2).1 As Figure 3 illustrates, this large market is diversified across many different industries.

1 Since 1993, the compound annual growth rate of institutional tranches has been 48%.

What is a leveraged loan?

In 2003, the new issue leveraged loan market

was larger than the primary high-yield

bond market.

Institutional investors have driven growth in the

leveraged loan market.

Institutional loans reach a record level in 2003.



Figure 2. Par Amount of Outstanding Institutional Leveraged Loans, 1996–2003 (Dollars in Billions)

13.6

34.7

73.3

101.1

117.3129.9 132.5

148.1

0

40

80

120

$160

1996 1997 1998 1999 2000 2001 2002 2003

Majority are Institutional Tranches


Figure 3. Par Amount of Outstanding Leveraged Loans by Industry, 2003

Chemical/Plastics 4%

Containers & Glass 6%

Ecological Services & Equipment 3%

Electronics/Electric 2%

Food Products 4%

Cable TV 6%

Business Equipment & Services 2%

Automotive 4%

Healthcare 9%

Hotels & Casinos 3%Industrial Equipment 3%

Leisure 4%

Publishing 7%

Telecommunications 9%

Utilities 4%

Other 30%


The growth in the institutional market has had many implications. For example, issuers and arranging banks structure deals to be more attractive to institutional investors, especially the larger issues. Furthermore, rating agencies, whose fortunes are tied to the needs of institutional investors, have dramatically increased the number of loans that they rate in the past few years. These positive trends, among others, should help overcome some of the obstacles that previously inhibited the growth in the market.

Broadening Investor Base Until the mid-1990s, the universe of loan investors included only banks and prime funds.2 Because these investors generally took positions in loans with the intention of holding until maturity, there was little need for a secondary market. Over the past few years, however, as more institutions realized that loan products could deliver

2 Prime rate funds are mutual funds that buy portions of corporate loans from banks and pass along interest designed to approximate the prime rate to shareholders.



high risk-adjusted returns, new types of investors, such as high-yield bond funds, hedge funds, insurance companies, and CDOs began to enter the market. Figure 4 illustrates the dramatic rise in the loan market share of CDO investors, in particular, during the past decade from only 4.2% to 66.7% in 2002. The current investor base has diverse needs, and many participants actively look for arbitrage opportunities and trade ideas in the secondary loan market. This impulse has helped fuel the secondary market and the dominance of the institutional loan tranche structure. We discuss the dynamic secondary loan market in the next section.

Figure 4. Primary Market for Institutional Loans by Investor Type, 1994 Versus 2002 Distribution of Investors in 1994 Distribution of Investors in 2002

Hedge & NY Funds1.1%

Finance Co. 7.6%CLO 4.2%

Insurance Company20.8%

Prime Rate Fund 75.1%

Prime Rate Fund20.2%

Insurance Company4.4%

CLO 66.7% Source: Standard & Poor’s.

Increasing Secondary Market Liquidity As mentioned, the number of institutional investors in the leveraged loan market has grown dramatically over the past decade. The rise in the number of participants (with varying investment vehicles and agendas) has had a positive impact on liquidity. As shown in Figure 5, from 1993 to 2003, the number of active institutional investment vehicles in leveraged loans increased from 14 to 372, and the annual volume of leveraged loans traded followed suit, climbing from approximately $15 billion to about $146 billion. Higher secondary trading volumes (see Figure 6) and smaller average trade sizes are direct indicators of the increasing vibrancy and liquidity of the secondary loan market. Ten years ago, the average trade size was $10 million. In 2003, the average trade size was approximately $1 million as loans traded readily among a variety of different types of counterparties.

Loan investors and secondary trading have

grown tremendously over the past decade.



Figure 5. Total Active Institutional Loan Investors, 1993–2003

14 1832

4664

122150

205 213

272

372

0

100

200

300

400

1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003


The industry’s trade association, The Loan Sales and Trading Association (LSTA) has contributed significantly to improved liquidity in the secondary loan market by standardizing documents, trade settlement time frames, and industry practices. The organization was founded by a group of banks in 1995 and has expanded to 143 members since then. The LSTA is open to participants in the loan market on both the buy side and the sell side. In 2003, the LSTA continued to implement a number of reforms to the secondary loan market including the following:

➤ New primary market standardized documents including credit agreements, bank books, amendments, and tax shelters.

➤ New secondary market standardized documents including distressed purchase and sale agreements, netting agreements, and trade criteria.

➤ Introduction of CUSIPs to the loan market for use as unique identifiers by agent banks, loan participants, and rating agencies. Use of CUSIPs is the critical first step in moving the loan market to an eventual straight-thru processing platform.

Figure 6. Annual Secondary Trading Volume, 1991–2003 (Dollars in Billions)

4 5 6 1326 33

5266 70 78 76

88

4 6 98

86

9

2442

48

58

65

912

0

25

50

75

100

125

$150

1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003

Distressed

Par


Standardization has improved.



Continuing Challenges to Loan Market Liquidity Despite dramatic improvements, liquidity in the loan market remains constrained for a few reasons, including minimum purchase sizes. A decade ago, $10 million was the industry standard trade size. This minimum requirement served as an impediment to secondary investors in the loan market. Although there is no official minimum amount that can be assigned, credit agreements can still have provisions that require this amount to exceed $5 million, or in some deals, even $10 million. However, the trend is positive: assignment minimums and trade sizes have been declining as a result of investor pressure. Investors are very sensitive to the relative weights of loan exposures in their portfolios, so they look to the secondary market as a way to balance this exposure. As shown in Figure 7, average assignment minimums dropped to $1.24 million in 2003.

Figure 7. Average Institutional Assignment Minimum for Loans of $100 Million or More (Dollars in Millions)

$6.25

$5.30 $5.10 $4.91$4.40

$2.35

$1.48$1.15 $1.24

0

2

4

6

$8

1995 1996 (40) 1997 (88) 1998(144)

1999(165)

2000(139)

2001(123)

2002(135)

2003(195)


In addition to minimum purchase sizes, high transfer fees have limited liquidity. The lead administrative agent bank charges a fee for each trade as compensation for keeping track of holders for documentation purposes. Although certain banks in the industry are trying to lower the cost of these fees to promote greater liquidity in the market, others are not, and as Figure 8 illustrates, average fees have remained in the range of $3,000–$3,500 per assignment since the mid-1990s.

Minimum purchase sizes and high trading fees affect liquidity in

the loan market.



Figure 8. Average Institutional Assignment Fee for Loans of $100 Million or More, 1995–4Q 03 (Dollars in Millions)

0

1,000

2,000

3,000

$4,000

1995 (16) 1996 (40) 1997 (88) 1998(144)

1999(165)

2000(139)

2001(123)

2002(135)

2003(195)

4Q03 (44)


Although loan market liquidity clearly remains a challenge for investors, it has improved enormously from only a few years ago. As loan investors continue to increase the pressure, liquidity will rise. We expect secondary loan market liquidity to approach that of the high-yield bond market as documentation and trade procedures standardize, transaction fees are eliminated, and minimum assignments are reduced.

Outlook: improved liquidity



Key Loan Characteristics

Floating-Rate Coupon Leveraged loans pay interest on a floating-rate basis, so interest payments on loans increase as market interest rates rise. This floating-rate structure is created by setting the interest rate of a leveraged loan at a spread above a benchmark market floating interest rate. The most commonly used benchmark in the leveraged loan market is LIBOR (the London Interbank Borrowing Overnight Rate). So a loan paying 3.0% above LIBOR (or L+300bp) would yield 4.2% annually if LIBOR were at 1.2% for a given period. As LIBOR moves, the interest payments of a leveraged loan will move with it.

This floating-rate coupon is one of the most significant differences between leveraged loans and high-yield bonds. Because high-yield bonds pay a fixed interest rate using a US Treasury bond benchmark, investors are exposed to movements in interest rates. If market interest rates rise, the fixed-rate high-yield bond will continue to pay the same lower interest rate. While derivatives can be used to hedge away this risk, this can only be done at a cost that cuts into expected returns. By contrast, the floating-rate interest payments of loans move with the market.

Maturity Term loans generally mature in five to eight years from the time of issue, a considerably shorter period than the ten-year average high yield bond maturity. In 2003, term loans had average maturities of approximately 5.6 years, as shown in Figure 9.

Figure 9. Average Weighted Term of Institutional Loans, 1998–2003 (Years)

7.16.9 6.7

6.15.8 5.6

2

4

6

8

1998 1999 2000 2001 2002 2003

Year

s


Leveraged loans pay a floating interest rate,

usually set at a spread above LIBOR.

Term loans usually mature in five to eight years.



Callability Loans are generally callable at par without penalty, meaning that the issuer can repay their loans partially or in total at any time. This structure differs from that of high-yield bonds, which are usually structured with a noncall period of three to five years. Occasionally, loans will have noncall periods or call protection that requires the issuer to pay a penalty premium for prepaying loans. These features are usually only added to loans in the primary market when investor demand is weak and a loan needs additional incentives to attract sufficient buyers. Figure 10 shows the 12-month rolling prepayment rate for loans at approximately 15% in 2003.

Figure 10. Repayment Rates, 1Q 97–4Q 03

0%

3%

6%

9%

12%

15%

18%

1Q 97 3Q 97 1Q 98 3Q 98 1Q 99 3Q 99 1Q 00 3Q 00 1Q 01 3Q 01 1Q 02 3Q 02 1Q 03 3Q 03


Covenants Loan facilities are structured with covenant tests that limit a borrower’s ability to increase credit risk beyond certain specific parameters. Covenants are outlined in the legal credit agreement of a loan facility that is executed at the time that a loan is issued. Typically, covenants are tested every quarter, and results are sent to all of the members of the bank group. Covenant tests provide lenders with a more detailed view of the credit health of a borrower and allow lenders to take action in the event that a borrower gets into credit trouble. A credit agreement for a leveraged loan may generally have between two and six covenants depending on the credit risk of the borrower and market conditions. Some commonly used covenants include:

➤ Minimum EBITDA

➤ Total leverage debt/EBITDA

➤ Senior leverage senior debt/EBITDA

➤ Minimum net worth

➤ Maximum capital expenditures

➤ Minimum interest coverage EBITDA/interest

Loans are typically callable and can be

repaid by the issuer at any time.

Covenants provide lenders with more

control over a borrower.



Covenants on leveraged loan facilities improve recovery because they allow lenders to limit credit risks such as, but not limited to, capital expenditures, leverage, and acquisitions. Covenants also allow lenders to have an early look at an issuer’s credit problems, often before the rest of the market, as an issuer must amend or repay its loans when covenants are breached. This amendment process allows lenders to improve their control and security interest in a troubled issuer, raising potential recovery values.

Ratings In response to strong investor demand over the past decade, the major debt rating agencies have dramatically increased the number of leveraged loan issuers that they rate. Moody’s, S&P, and Fitch all actively rate and monitor loan deals. The number of rated loans has soared to the point where now 70% of all new issues receive a rating from at least one agency.

Although methodologies used to determine the ratings differ somewhat from agency to agency, significant progress in refining the methodology has occurred across all agencies, meaning that investors are receiving more accurate information on a wider number of loans, and this should allow for more reliable pricing. Such information is especially important given the material rise in secondary trading and the corresponding entry into the market of a large number of investors that have to carry out regular mark-to-market portfolio pricing. This phenomenon has transformed the leveraged loan market so that a rating change by any one of the agencies can cause a significant change in the value of a loan. When this type of price swing occurs, these mark-to-market investors have to rearrange their portfolios, creating arbitrage opportunities for investors that can act quickly to take advantage of these opportunities.

The ratings given to loans are primarily based on two factors:

1 Probability of default

2 Expected recovery rate

The probability of default for a bank loan is approximately equal to that of a bond. Despite this fact, a loan is frequently given a higher rating than a bond of similar size and duration. The reason for this disparity lies in the fact that default rates do not capture a critical, value-adding component of a loan — its higher status in the capital structure of a firm relative to a bond. Because a bank loan is generally a senior secured debt obligation, the average recovery rate for loans is significantly higher than that for bonds, which, at best, tend to be senior unsecured debt obligations.

The widespread rating of loans is a relatively recent phenomenon that did not take off until the mid-1990s. In the past, when loans were not rated, market participants generally estimated that the loan should be one notch up from the most senior unsecured bond. While this rule is fairly accurate on average, it has some serious shortcomings when used to evaluate pricing for individual credits. In fact, according to a 1998 study carried out by Moody’s, only 37% of loans were actually rated exactly one notch higher than the senior unsecured bond. This means that an investor exclusively using this “one-notch” rule to price the premium paid on a loan’s higher recovery rate would have mispriced the loan more than 60% of the time.

Covenants improve recovery on

leveraged loans.

Ratings agencies have increased their coverage

of leveraged loans.

Leverage loans usually have higher ratings

than high-yield bonds.



This is not to say that the one-notch rule does not have practical uses. It is still useful as a benchmark from which to start one’s credit analysis. As a tool to price loans, however, it is clearly inadequate. Instead, investors need to follow the lead of the rating agencies and look very carefully at the credit’s attributes to determine how such factors as industry, corporate structure, legal subordination, underlying collateral quality, and a host of other factors will affect recovery rates in cases of default, because these factors can cause the recovery rates of seemingly similar loans to differ significantly.

Security Leveraged loans are generally structured with a lien against the assets of the borrower. These asset claims are also known as the security of the loan. Secured loans have a number of advantages over unsecured parts of a company’s capital structure. In the event of a default, the lenders can take possession of the borrower’s assets to which they have a claim and sell them or operate them for cash. The position of a debt instrument in the firm’s capital structure and the degree to which the debt is backed by liquid assets are important indicators of expected recovery rates.

Recovery rates on defaulted loans are consistently higher than recoveries on unsecured parts of the capital structure. As we discussed, the higher recovery rates are primarily due to the senior position of leveraged loans in an issuer’s capital structure and the security interest that loan holders have in an issuer’s assets.

Fees The fees for leveraged loans generally comprise “up-front fees” and “commitment fees” that vary with market conditions. Figure 11 shows historical fee levels for the past nine years.

Figure 11. Average Total Fee, 1995–2003 (In Basis Points)

10

20

30

40

50

60

70

80

1995(29/8)

1996(56/20)

1997(62/38)

1998(210/107)

1999(202/145)

2000(212/127)

2001(203/110)

2002(219/196)

2003(180/339)

Basi

s Po

ints

Pro Rata Institutional

Note: Numbers in parentheses represent the number of pro rata and institutional issuers. Source: Standard & Poor’s.

The one-notch rule can be used for

benchmarking rather than pricing.

Bank loans are senior secured debt.



Up-Front Fees

In the leveraged loan market, issuers pay one-time “up-front” fees at closing to attract banks and institutional investors to invest in their loans. Up-front fees for pro rata loans tend to be higher than up-front fees on institutional term loans because of their lower coupons than institutional tranches and less favorable supply/demand conditions.

Commitment Fees

Also called unused fees, commitment fees are assessed on an ongoing basis on the committed but undrawn component of a revolving credit facility. This fee accrues daily on the undrawn balance of a loan at an annualized interest rate specified in the credit agreement. As banks evolve from lending based on relationships to lending based on returns, the average commitment fee has breached the 50bp level, a benchmark that was once considered a ceiling on these fees. This trend is expected to persist in the future, as retail bank investors continue to raise the bar on returns for the pro rata tranches of leveraged loans.

Loan Structures A corporate leveraged loan generally has multiple tranches:

➤ The revolving credit facility

➤ The term loan A

➤ A single or multiple institutional term loans (B, C, D . . .)

Revolving Credit Facilities

The revolving credit facility is an unfunded or partially funded commitment by lenders that can be drawn and repaid at the issuer’s discretion until maturity. The borrower pays a nominal commitment fee on the undrawn amount (usually 50bp) and a coupon on the drawn amount (usually the same coupon as the term loan A). Maturities are usually either one year (bridge facility or short-term debt 364-day facility) or in the three- to five-year range. Revolvers generally serve as liquidity facilities for borrowers and can be drawn at any time for operational purposes, seasonal capital needs, or letter of credit issuance, as outlined by the credit agreement. From an investor’s perspective, the uncertain funding requirements and interest payments make revolvers difficult to administrate and fund.

Amortizing Term Loans

A term loan A (TLA) is a fully funded term loan that usually amortizes throughout the life of the loan but can also be structured with a bullet amortization. Unlike the revolving credit facility, once the borrowed amount is paid back to the lenders, the borrower cannot reborrow the money under the TLA facility. The TLA and the revolving credit facilities are bundled together during original syndication, but can be unbundled for subsequent redistribution in the secondary market. In virtually every case, the spread and term of the revolving credit facility and TLA are the same.

Pro rata fees tend to be higher than fees on institutional loans.

A leveraged loan facility is typically

made up of three parts.

Revolvers can be drawn down and repaid, providing flexible

liquidity for the issuer.

Term loan A facilities amortize more quickly,

ensuring debt



Institutional Term Loans

Institutional term loans represent the most liquid category of corporate leveraged loans. Unlike the TLA, amortization payments of institutional term loans are usually more heavily backloaded or come as a bullet payment at final maturity, as is the case for many bonds. For example, on a six-year institutional term loan, the payments over the first five years may be only on interest, with entire principal payments made over the course of the final year. Institutional investors favor institutional term loans because of their more predictable funding requirements, maturities, and interest income streams. Institutional tranches are usually named in alphabetical order, term loan B, C, D, etc., depending on the number of tranches. For example, term loan A might have a maturity of five years, term loan B six years, term loan C seven years, etc. In most leveraged loans, investors expect to receive an additional 25bp–75bp in coupon for each additional year until maturity, although in practice spreads reflect what the market is willing to price in compensation for a longer maturity. Recently banks have taken greater interest in buying institutional tranches in the primary market, joining insurance companies, hedge funds, CDOs, and other institutions seeking the higher coupons available on institutional term loan tranches.

Term loan B facilities are fully drawn and

structured for investors’ needs.



Pro Rata Loans

Overview

Structure of Pro Rata Loans

The pro rata portion of a corporate bank debt facility is the traditional loan structure, historically syndicated and held almost entirely by banks. It usually comprises two pieces, a revolving credit facility and a TLA tranche (see “Loan Structures”). This segment of the loan is called the “pro rata” portion because banks that take part in the syndication must commit to an equivalent proportion of both the revolving credit facility and the term loan A. Coupons (shown as LIBOR plus a spread) on pro rata tranches are often lower than comparable coupons on institutional tranches because of higher up-front fees and accelerated payments associated with pro rata tranches. Because of the funding requirement for the revolver and the accelerated amortization of the TLA, the universe of lenders tends to be restricted to traditional corporate lending banks.

Pro Rata Market

At $75 billion in 2003, new issuance in the pro rata loan market was essentially flat with the $81 billion issued in 2002. The pro rata loan market is comparable in size to the high-yield bond market, which produced $137 billion in new issue in 2003 and $65 billion in 2002. The large size of the pro rata loan market means that investors have opportunities to invest across a wide spectrum of industries and credits and to benefit from the relative value relationships between pro rata loans and other asset classes.

Pro Rata CLOs

Higher recoveries and discount pricing, along with the overall depth of the market, make pro rata loans an attractive option for CLO collateral. The major challenge for the CLO market was to create an appropriate structure to handle the funding risk of the revolving part of the loan. In 2003, the very first CLO transaction backed by pro rata loans was brought to the market. That transaction used an innovative synthetic structure to separate the funding and credit risks of the revolvers, capitalizing on the high ratings of the underwriting bank. The funding risk stayed with the bank, whereas the credit risk was sold into the CLO structure. This was the first transaction that provided CLO investors with wide access to the broad pro rata loan market.

Key Characteristics

Discount

Pro rata loans generally trade at a discount to par. The discount is due to the combination of a lower coupon for the pro rata loan relative to institutional term loans, certain impediments to liquidity, and the selling pressure caused by banks seeking to rationalize their balance sheets and credit exposures. Pro rata loans are attractive for buyers, including structured vehicles, that may benefit from purchasing discounted instruments.

Pro rata loans consist of a TLA and a revolver.

In 2003, pro rata loans accounted for 45% of

the new issue leveraged loan market.



Security

Pro rata loans, like institutional term loans, are senior secured and have maintenance financial covenants. Covenants give investors an early “seat at the table” in the event that the issuer’s credit deteriorates, and lenders often use them to improve their position in terms of security, collateral, coupon, or fees.

Prepayment Rate

Because of amortization, prepayments, refinancings, and corporate events such as assets sales and mergers or acquisitions, pro rata loans tend to be repaid prior to their scheduled maturity. Increases in the prepayment rate garner yield windfalls because of quicker-than-expected recovery of purchase price discounts. Prepayment benefits holders of pro rata loans more than holders of institutional term loans because pro rata loans are generally bought at deeper discounts to par.

Superior Recovery

Losses in the event of default are generally lower for pro rata asset classes than for institutional term loans, because the revolving credit portions of pro rata loans, on average, are not fully drawn at default. The obligation to fund the undrawn portion of a revolving credit facility ceases upon default, and thus creates an effective windfall (i.e., tantamount to a repayment of that portion of the pro rata loan at par).

Barriers to Entry Pro rata loans represent a robust yield/value opportunity. The potential price arbitrage versus institutional term loans will likely be maintained because of credit agreements that restrict ownership to banks or other holders the issuer finds acceptable. The requirement for borrowers to consent to transfers of pro rata loans means that the pool of acceptable counterparties is likely to grow slowly. An agent bank can facilitate this approval process but may require cash collateral to do so. Similarly, lenders/investors must be capable of properly managing the variable funding requirements of revolver borrowing. This restricts the number of investors who can buy pro rata loans.

Investment Opportunities Revolvers and TLAs are attractive from a relative value perspective as they generally trade at a discount to par in the current market, while institutional term loans generally trade at or above par. Because the majority of corporate leveraged loans can be repaid at par (100 cents on the dollar) at any time without a penalty, investors who buy below par collect the difference as a gain upon refinancing. A number of market forces have driven this relationship, including short new issue supply in the loan market, relatively fewer pro rata lenders, the rise in the number of institutional loan investors, and a strong demand for product that has focused on TLBs.

The fact that few investors are able or willing to participate in the pro rata loan market creates additional investment opportunities. The majority of pro rata loan holders are relationship banks that buy the loans during original syndication. Banks continually readjust their balance sheet strategies to diversify credit risk and free up capital for new deals, creating ongoing opportunities to buy pro rata loans in the secondary market at a discount. As the number of pro rata lending banks decreases, we expect this market dynamic to persist for the foreseeable future.

Pro rata loans usually trade at discount.



Middle-Market Loans

Overview The middle-market segment of the leveraged loan market generally comprises smaller companies that satisfy the following criteria: (1) less than $500 million in revenues and $50 million in EBITDA; and (2) a loan facility smaller than $150 million in total size. Within those criteria, large middle-market loans have deal sizes of $100–150 million and issuers with EBITDA of $25-50 million, while standard middle-market loans are smaller. In 2003, 79 middle-market loans were issued with a total size of $22.9 billion, up considerably from $13.6 billion in 2002. Middle-market loan lenders vary across different industrial sectors, with the healthcare, services/retail, industrial, and media sectors accounting for 62% of the total market volume in 2003 (see Figure 12).

Figure 12. Loan Volume by Broad Industry Classification for Deals With $50 Million or Less of EBITDA, 2003

Healthcare 17%

Services/Retail 17%

Industrial 15%Media 13%

Entertainment& Leisure 11%

Computers &Electronics 5%

Food & Beverage 5%

Building Materials 4%

Utilities 3%

Other 10%


Key Characteristics

Liquidity

Middle-market loans are generally less liquid than leveraged loans because of their smaller size and lower visibility among the institutional investor community. This tighter liquidity makes middle-market loans more suitable for buy-and-hold investors who want to collect the higher yield and tend not to trade as actively. Many CLO structures could benefit from holding discounted middle-market loans and collecting the higher associated yields, because they do not require the liquidity to trade the loans actively. Middle-market institutional loans pay a coupon of approximately L+400bp, nearly 150bp more than the coupon on a comparable BB/BB- rated institutional leveraged loan (see Figure 13).

Middle-market loans are generally issued by smaller

companies than typical leveraged loan issuers.

The secondary market for middle-market

loans is less liquid.



Figure 13. Average Weighted Institutional Spreads: BB/BB- Leveraged Loans Versus Middle-Market Loans, 1Q 99–4Q 03

L+150

L+250

L+350

L+450

L+550

1Q 99 3Q 99 1Q 00 3Q 00 1Q 01 3Q 01 1Q 02 3Q 02 1Q 03 3Q 03

Leveraged Loans Middle Mkt Loans


Leverage

Leverage on middle-market loans can be slightly higher than that for comparable leveraged loans, but is currently very similar at an average of less than 4.0 times debt/EBITDA compared with the 4.0 times average for the overall leveraged loan market. Figure 14 illustrates historical middle-market loan leverage, and Figure 15 shows the same for the overall loan market. In general, middle-market loans tend to include more secured bank debt, so senior secured leverage is often higher than in the leveraged loan market, as evidenced by senior leverage of 3.0 times for middle-market loans versus a much lower 2.3 times for the overall loan market.

Figure 14. Rolling Three-Month Debt/EBITDA and Senior Debt/EBITDA Ratios for Issuers With EBITDA of $50 Million or Less, Mar 99–Dec 03

1.0x

1.5x

2.0x

2.5x

3.0x

3.5x

4.0x

4.5x

Mar

99

(85)

Jun

99 (1

13)

Sep

99 (1

00)

Dec

99 (7

0)

Mar

00

(82)

Jun

00 (9

1)

Sep

00 (7

9)

Dec

00 (3

7)

Mar

01

(41)

Jun

01 (3

9)

Sep

01 (2

2)

Dec

01 (2

1)

Mar

02

(35)

Jun

02 (2

6)

Sep

02 (2

2)

Dec

02 (4

6)

Mar

03

(21)

Jun

03 (1

8)

Sep

03 (1

7)

Dec

03 (2

3)

Total Debt/EBITDA Senior Debt/EBITDA

Note: The numbers in parentheses represent the number of issuers. Source: Standard & Poor’s.

Leverage can be higher on middle-market loans.



Figure 15. Average Debt Multiples of Highly Leveraged Loans, 1987–2003

8.8

7.16.7

5.35.0 5.1 5.3 5.2

5.8 5.75.2

4.54.0

3.7 3.8 4.0

2.3

4.7

3.7 3.4 3.42.6 2.7 2.8

3.3 3.5 3.6 3.5 3.32.9

2.2 2.4

0.0

2.0

4.0

6.0

8.0

10.0

1987 1988 1989 1990 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003

Total Debt/EBITDA Senior Sec Debt/EBITDA


Ratings

Middle-market loans are frequently not rated by the major debt rating agencies like Moody’s and S&P. In 2003, only 17% of middle-market loans issued had debt ratings (see Figure 16). The primary reasons for the lower number of ratings are weaker investor demand and rating fees. Investors tend to focus on larger, widely syndicated leveraged loan deals with debt ratings that allow them to match loan investments to the requirements of their investment vehicles. In addition, the smaller size of a middle-market deal makes it more difficult to justify paying the fees (which can amount hundreds of thousands of dollars) required to obtain a debt rating. As a result, the primary lenders/investors in the middle market are commercial banks and finance companies that do not require debt ratings for their lending process (see Figure 17). Moreover, these institutions tend to view their middle-market loan activity as a part of a larger overall business relationship with the issuing company. Institutional investors have just begun to increase exposure to middle-market loans in the past year because of a lack of standard leveraged loan paper, but they remain a small portion of the overall market.

Figure 16. Loan Volume by Rating for Deals by Issuers With EBITDA of $50 Million or Less (Total New-Issue Volume: $10.0 Billion), 2003

NR83%

BB/BB-6%

B+/B11%

NR Not rated. Source: Standard & Poor’s.

Middle-market loans typically do not have

debt ratings.



Figure 17. Primary Markets for Highly Leveraged Loans for Issuers With EBITDA of $50 Million or Less, 1997–2003

0%

25%

50%

75%

100%

1997 1998 1999 2000 2001 2002 2003

Shar

e of

Mar

ket

Domestic Bank Finance Co. Foreign Bank II Securities Firm


Investment Opportunities Middle-market loans offer investors credit exposure to a pool of issuers beyond those found in the broadly syndicated leveraged loan and high-yield bond markets. Middle-market loans can provide higher yields and attractive price discounts for investors who are willing, in some cases, to take on additional credit risk. The potential for higher yield versus institutional term loans will likely persist because of significant structural differences between the two classes that result in a limited lending group for middle-market issuers. Besides potentially higher yields, middle-market loans, like institutional term loans, are senior secured and have maintenance financial covenants. We think that it is reasonable to expect that middle-market loan recoveries should be roughly comparable to institutional term loans because both asset classes have a senior secured claim on the assets of the issuer.



European Leveraged Loans

Overview The European leveraged loan market provides additional opportunities for investors within the senior secured loan asset category. Annual issuance for this market reached a record €78.5 billion in 2001. Issuance then declined to €40.1 billion in 2002 and climbed to €41.8 billion in the first ten months of 2003 (see Figure 18).

Figure 18. European Leveraged Loan Volume, 2000–Oct 03 (Euro in Billions)

59.5

78.5

40.1 41.8

0B

20B

40B

60B

80B

12000 2001 2002 Jan–Oct 2003


The new issuance was diversified across industries and countries. In 2003, four countries, the United Kingdom, France, Italy, and Germany, accounted for 75.4% of the total new issuance, with the UK constituting a little less than one-third. Lending was well balanced across a number of different industrial sectors, with the printing and publishing sector leading with 17.2% of market share (see Figure 19).

Figure 19. European Leveraged Loan Volume by Country and Industry Jan–Oct 03 Jan–Oct 03

Other 39.7%UK 31%

Printing & Publishing 17.2%

Gaming & Hotel 4.4%Telecom 4.7%

Services 5.5%Retail 5.7%

Media 6.0%

Food & Beverage 6.2%

Building Materials 10.6%

France 21%

Other 8%Denmark 2%Spain 3%

Ireland 4%

Netherlands 7%

Germany 12%

Italy 12%




European banks have traditionally dominated the primary issue leveraged loan market, with almost two-thirds of the total market volume (see Figure 20). Institutional investors (mostly, CLOs) accounted for only 20% of the total size of the European leveraged loan market. These findings are in sharp contrast with the United States where institutional investors dominate the primary loan market.

Figure 20. Primary Market for European Leveraged Loans by Investor Type,a Latest 12 Months as of 31 Mar 03

European Bank 66%

Canadian Bank 2%Asian Bank 4%

Insurance 1%

Finance Co. 1%

MDO Managers 20%

US Bank 3%

Securities Firm 3%

a Excludes US dollar tranches. Source: Standard & Poor’s.

European Mezzanine Bank Loans A growing part of the European loan market is the European mezzanine market (see Figure 21 for the historical growth of the market), which has now become the primary source of funding for European leveraged buyouts. A European mezzanine loan is generally a subordinated second secured debt obligation.

Figure 21. European Mezzanine Market Evolution (Euro in Millions)

0

500

1,000

1,500

2,000

2,500

3,000

3,500

4,000

1998 1999 2000 2001 2002 2003

(EUR

in M

illio

ns)

Source: Fitch.

European mezzanine bank debt is a floating-rate instrument, and its coupon usually includes cash and a pay-in-kind (PIK) component. Figure 22 shows a typical mezzanine structure and compares it with other debt instruments. Investors are attracted by the higher spread relative to senior secured loans and greater protection than offered by high yield bonds, achieved through covenants and security. However, European mezzanines still largely constitute a privately rated asset class, and only limited performance data are available.



Figure 22. A Comparison of Typical European Funding Structure Characteristics Senior Debt Mezzanine High Yield Equity Security Yes — First Ranking Yes — Second Ranking Usually None None

Ranking Senior Contractually Structurally Junior Subordinated Subordinated

Covenants Generally Comprehensive Often Track Senior Debt Covenants

Less Restrictive. Mostly Financial.

None

Term 5-9 Years 6-10 Years 7-10 Years Open Ended

Income Cash Pay — Floating Cash Pay — Floating Cash Pay — Fixed Dividends — Uncertain, Usually Cash

Pay

Source: Fitch.

Key Characteristics

Leverage

European leveraged loans exhibit debt multipliers similar to those of US leveraged loans. Figure 23 shows debt-to-EBITDA ratios for the European loan market.

Figure 23. Average European Leverage Statistics — Rolling Three-Month Debt Multipliers, Jan 98–Oct 03

0.0x

1.0x

2.0x

3.0x

4.0x

5.0x

6.0x

Jan 98 Jul 98 Jan 99 Jul 99 Jan 00 Jul 00 Jan 01 Jul 01 Jan 02 Jul 02 Jan 03 Jul 03

Senior Sec Debt/EBITDA Total Debt/EBITDA


Recoveries

Recovery data for European loans is limited, given the private nature of the loan market. In 2000, Fitch rating agency conducted research on an unnamed basis that showed an average recovery rate of 76.5%, which is slightly lower than the correspondent recovery rate for US loans. However, these findings were hampered by the small size of the data sample used in the analysis. In addition to the limited data availability, differences exist in insolvency regimes across Europe. Because of these differences, recovery rates are expected to vary across European countries depending on the jurisdiction.



Primary Spreads

Figure 24 provides comparison of the historical spreads on institutional BB/BB- rated new issues in the United States and Europe. The historical data show that European spreads were generally lower than US spreads between 1999 and 2002. However, by the beginning of 2003, US spreads were rallying significantly, and European spreads ended up on top of their US counterparts.

Figure 24. Weighted-Average New Issue Spread of European and US BB/BB- Institutional Issuers, 2H 99–3Q 03

L+150

L+200

L+250

L+300

L+350

L+400

0 2 4 6 8 10 12

EUROPE

US


Investment Opportunities Investing in European leveraged loans can provide additional geographic and issuer diversification opportunities beyond the US obligors. Although recovery data are limited, a European leveraged loan represents the senior secured debt of an issuer, and therefore, it is expected to have significant recoveries in case of default.



Collateralized Loan Obligations

Efficient Access to Loan Market Investment Opportunities — Introducing CLOs As we have discussed in this report, leveraged loans represent a broadly diversified, rapidly growing market. The unique characteristics of loans, such as high recoveries and stable prices, appeal to many investor types (including CLOs), especially during events such as the most recent bear credit market.

Despite its unprecedented growth, the loan market investor base has yet to reach the scale of the high yield bond market. Participation in CLOs allows investors to capitalize on existing price inefficiencies in the loan market, diversify their exposure to the bank loans and utilize professional management expertise and resources. Various types of investors, ranging from banks to high net worth individuals, have used CLOs to gain leveraged exposure to bank loans (see Figure 25).

Figure 25. CLO Investor Profile Senior/Subordinate (AAA/AA/A/BBB) Securities Banks Insurance Companies Conduits Fund Managers

Mezzanine (BBB/BB) Securities Insurance Companies Banks (Specialized Funds) Hedge Funds Fund Managers

Income Notes (CDO Equity) Insurance Companies Banks High Net Worth Individuals Alternative Investment Group/Special Investment Groups

Source: Citigroup.

Basic CLO Structure CLOs are created by applying asset-backed structuring technology to a pool of bank loans. The formation of a CLO begins with the establishment of a special purpose vehicle (SPV) to acquire a pool of bank loans (see Figure 26). The average collateral pool size is usually between $300 million and $500 million par value with the total exposure diversified across 100–200 distinct obligors in 20–30 industries.

CLOs allow investors to gain leveraged exposure

to the loan market.



Figure 26. A Typical CLO Structure

SPV("Issuer")

Portfolio

Interest + Principal

Manager

Class B

Class A

Class D

Class C

Equity

PortfolioManagement

Proceeds

Proceeds

Residual Cash

Proceeds

BankCounterparty

HedgeAgreements

LIBOR + Spread+ Principal

Source: Citigroup.

To fund the acquisition of the debt obligations, the SPV issues rated and unrated liabilities (tranches). The expected average lives of these CDO liabilities range from 6 to 12 years, depending on the tranche’s seniority. Because the majority of these liabilities are highly rated, the CLO can raise most of its capital cheaply in the investment-grade market and invest it more profitably in the leveraged loan market.

A typical CLO consists of five to seven rated tranches with the ratings ranging from AAA to BB and unrated income notes (also known as the equity tranche). The desired tranche ratings are achieved through obligor and sector diversification and leverage, and by employing the payment distribution waterfall designed to protect the more senior note holders of the deal’s liability structure.

Figure 27. A Simplified Waterfall

Interest & Principalfrom Collateral Pool

Manager orServicer

Senior Fee

Class A NoteInterest

Class A NotePrincipal

Class ACoverage Tests

Fail Fail Fail

Pass

Class B NoteInterest

Class B NotePrincipal

Class BCoverage Tests

Class C NoteInterest

Class C NotePrincipal

Class CCoverage TestsPass

SubordinatedManagement Fees

Equity

Redeem Most SeniorOutstanding

Class of Notes

Pass

a a

a Subject to delevering of the more senior tranches. Source: Citigroup.

A CLO consists of rated tranches and

income notes.



The waterfall directs proceeds from the underlying collateral pool to the liability note holders, ensuring higher asset coverage for the senior tranches (see Figure 27). Principal and interest cash flow is paid sequentially from the highest-rated class to the lowest. However, if the cash flow is insufficient to meet rated note costs or certain asset coverage tests are not met, most or all cash flow is diverted from the equity tranche and paid to the most senior tranche. The asset coverage tests are divided into two groups: overcollateralization (OC) and interest coverage (IC) tests. The former measures the amount of debt/asset coverage for a tranche, while the latter evaluates available interest proceeds to make coupon payments on the liability tranches.

A CLO investor can achieve a targeted return/risk profile by choosing a particular tranche in which to invest. The coupon margin reflects the relative riskiness of the tranche, and it increases with the lower ratings of the notes. Figure 28 shows a sample CLO capital structure. The income notes represent the riskiest investment, and therefore, they offer the highest potential return to compensate for this exposure. These notes receive the residual interest cash flow remaining after payment of fees, rated note holder coupons, and the satisfaction of any asset coverage tests. Depending on their risk/return objectives, investors can position themselves across the capital structure of a CLO.

Figure 28. Sample CLO — Capital Structure Assets Average S&P Rating of the Collateral Loans B+ Principal Amount (MM) $325.0

Liabilities Class A Class B Class C Class D EquityS&P Rating AAA A- BBB BB NRMoody’s Rating Aaa A3 Baa2 Ba2 NRPrincipal (MM) $245.0 $33.0 $13.0 $14.8 $25.5Percentage of Capital Structure 74.0% 10.0% 3.9% 4.5% 7.7%Stated Final Maturity (Years) 12.0 12.0 12.0 12.0 12.0Average Life (Years) 6.3 8.5 9.1 9.6 —

NR Not rated. Source: Citigroup.

CLO Asset Manager3 Once a CLO is issued, the collateral manager manages the portfolio according to the investment guidelines set forth in the bond indenture and within parameters necessary to satisfy the rating agencies. Within these guidelines, the manager sells and buys assets and, during the reinvestment period, reinvests collateral principal cash flows into new loans. The investment guidelines typically require that the CLO manager maintain a minimum average rating and portfolio diversity with the goal of muting any adverse effects that trading activity may have on note holders. The primary responsibility of the CLO collateral manager is to manage the portfolio in a way that minimizes losses to the note holders stemming from defaults and discounted sales. To this end, all note holders rely on the manager’s ability to identify and retain creditworthy investments. In particular, income note holders are substantially

3 For more details, please see The ABCs of CDOs Equity, Citigroup, July 2000.

Cash flow is paid sequentially, starting with the most senior

CLO classes.

CLO income notes are typically unrated, and

they represent the most subordinated part of the

CLO capital structure.

Trading decisions can have a significant effect

on the CLO performance.



dependent on a manager’s performance; the initial asset selection and trading activity throughout the reinvestment period are critical to achieving high returns.

Because note holder returns hinge upon good collateral manager performance, the choice of a CLO manager is a crucial decision for the investors. When choosing the collateral manager, the following key attributes should be examined in depth:

➤ The track record managing loan portfolios.

➤ Experience managing within the CLO framework.

➤ Level of institutional support.

➤ Investment and trading philosophy.

➤ Expertise in each asset class that the manager is permitted to invest in.

➤ Importance of CLO product to overall organization.

➤ Manager’s access to loans.

A manager with a deep understanding of the underlying credit fundamentals of the various loan markets can make informed, credit-based trading decisions, not trading decisions based on price movements.

CLO Market Today The CLO market is a subsector of the broader CDO market. The latter has grown dramatically since the middle of the 1990s, reaching $70 billion in new issuance in 2003, close to its record of $78 billion in 2001 (see Figure 29).

Figure 29. Global CDO Growth, 1995–2003 (Dollars in Billions)

0

10

20

30

40

50

60

70

80

$90

1995 1996 1997 1998 1999 2000 2001 2002 2003

Sources: Bloomberg, Creditflux, IFR Markets, MCM, Fitch, Moody's, and Standard & Poor’s.

As the overall CDO market has grown so has the CLO portion of that market. As Figure 30 illustrates, CLOs accounted for 30% of all US transactions rated by S&P in 2003, up by 6.2% since 1998. By contrast, high-yield bond transaction (high-yield CBO) issuance data show a sharp decline from approximately 33% in 1998 to just above 1% in 2003.

A good CLO manager bases trading decisions on credit

fundamentals rather than price movements.

In 2003, the global CDO market has reached $70

billion in new issue.

CLO market share has grown with the market.



Figure 30. CDO Collateral — Distribution of US Cash Flow CDOs, 1998 Versus 2003 1998 2003

Trust Preferred 11.4% Arbitrage CLO 30.0%

Arbitrage HY CBO 1.4%Balance Sheet CLO 2.1%

CDO of ABS 24.3%CDO of CDO 1.4%Distressed Debt 2.1%

Real Estate CBO 8.6%

Retranching 18.6%

Arbitrage CLO 23.8%

Retranching 1.6%Project Finance 1.6%

Master Trust CLO 17.5%

EMCBO 19.0%

Balance Sheet CLO 1.6%Arbitrage IG CBO 1.6% Arbitrage HY CBO 33.3%


The recent credit market blow-ups and soaring default rates of 2000-2002 confirmed the resilience of leveraged loan collateral, steering more investors towards CLOs and away from traditional high-yield CBOs. In 2003, S&P rated only two new high-yield CBOs as opposed to 42 new CLO transactions. Many CDO investors have moved from CBOs into CLOs, attracted by the higher stability and strong returns associated with CLOs. This has caused a surge in the overall demand for primary loan issues. S&P estimates that approximately 67% of all issues in the primary institutional loan market were placed into various CLO vehicles in 2002.4

Key Drivers of CLO Outperformance Recent rating agency data have revealed striking differences in the historical performance of CLOs and compared with high-yield CBOs. In particular, CLO tranche ratings have been much more stable than CBO tranches rating and corporate debt ratings. Figure 31 illustrates Moody’s rating transition rates in these three asset classes.5 Moreover, as evident from Figure 32, the severity of CLO downgrades was less pronounced than that for CBOs. The historical performance of CLOs indicates far fewer and less pronounced rating downgrades than for HY CBOs and corporate debt.

Figure 31. Moody’s Historical CLO and CBO Rating Downgrades, 1996–2002

0%

5%

10%

15%

20%

25%

Aaa Aa2 Baa2 Baa3 Ba3

CLO

CBO

Corporates

Sources: Moody’s Investors Service.

4 S&P estimated that about 56% of new issuance in the first three quarters of 2003 was placed into CLOs.

5 The S&P studies agree with the Moody’s findings. In fact, S&P did not downgrade a single CLO tranche in 2003.

CLOs exhibited greater ratings stability than the

alternative asset classes.



Figure 32. Maximum One-Year Historical Downgrades Lowest One-Year Rating Transition

Rating CBO CLO Aaa Ba3 Aa2 Aa2 Caa1 A1 A3 Caa3 Baa3 Baa2 Ca/C Ba2 Baa3 Ca/C B2 Ba3 Ca/C Ca/C

Source: Moody’s Investors Service.

In addition to the slower pace and scale of downgrades, most CLO downgrades were localized: according to a Moody’s 2003 study, the majority of CLO downgrades have been limited to a handful of CLO managers. In fact, 56% of all CLOs downgraded by the agency in 2002 were associated with just three collateral managers. The same study indicated that all downgrades in earlier years were associated with the same three managers.

Why are CLO ratings so stable? The answer can be found in two main areas: superior performance of loan collateral and CLO-specific collateral manager expertise. Broad obligor and sector diversification, floating-rate collateral, and high recovery and prepayment rates all augur well for CLOs. In addition, as we discuss below, the typical bank loan manger mentality is well suited to managing loan portfolios in the CLO context.

Recovery Rates Because they occupy the most senior part of an issuer’s capital structure and are secured by its assets, defaulted bank loans often have substantially higher recovery rates than the more subordinated debt obligations of an issuer (see Figure 33). Over the 1988–2003 period, senior secured bank loans had an average recovery of 77.5% of par value, while senior unsecured debt recovered 41.5% of par value and junior subordinated bonds recovered only 22%. Even in times of credit duress, such as the 1998–2002 period, loans realized an average recovery rate of 74.1%, while senior unsecured debt recovered less than half that amount (36.8%). High recovery rates make it less likely that the collateral backing the CLO will deteriorate sharply. The CLO manager receives significant principal proceeds from the defaulted assets and has the opportunity to reinvest the cash flow into new collateral.

Figure 33. Historical Recovery Rates, 1988–2003 Versus 1998-2002

0% 10% 20% 30% 40% 50% 60% 70% 80%

Loans

Senior Secured Notes

Senior Unsecured Notes

Senior Subordinated Notes

Subordinated Notes

Junior Subordinated Notes 1998-2002

1988-3Q 2003


Leveraged loan characteristics and

CLO manager expertise enhance the solid

performance of CLOs.

Leveraged loans maintain high recovery

levels even in a stressful economic environment.



Prepayment Rates

Most corporate bonds have covenants that govern an issuer’s prepayment/call rights. High-yield bond prepayments are typically restricted for the first three to five years. By contrast, the vast majority of bank loans can be prepaid at par at any time without penalty. In fact, the average leveraged loan prepayment rate during the recent low interest rate environment was approximately 20%–25%, easily topping the average 5%–8% call rate in the high-yield corporate market. For this reason, the CLO collateral manager usually has significant principal cash flows to reinvest in additional collateral assets or to de-lever the liabilities. This is particularly useful in a stressed economic environment with the downward credit pressure in the underlying portfolio.

Floating Interest Rates

Because most CDO liabilities bear floating-rate coupons (i.e., a spread over LIBOR), floating-rate leveraged loan collateral effectively eliminates the interest rate mismatch between the assets and liabilities of a CLO. By contrast, CDOs that are backed predominately by fixed-rate high-yield bonds must enter into an interest rate hedge agreement with a third party to hedge against interest rate risk. Typically, a high-yield CBO periodically pays the counterparty a predetermined fixed interest rate on a fixed notional amount and in turn it receives floating-rate payments (usually determined by the value of LIBOR on the preceding payment date). The hedge balance is typically structured to decrease over time to mimic the expected amortization schedule of the notes and, hence, reduce the risk of the transaction’s being overhedged. Hedge payments are more senior than liability payments in a payment waterfall structure and, thus, may have a pronounced effect on the overall performance of a high-yield CBO.

Although the outstanding balance of the hedge amortizes with the time, the amortization schedule is determined at inception and may differ significantly from the realized amortization of CDO liabilities. In particular, in a highly stressed credit environment (for example, 2000 to 2002), the most senior tranche of a CDO may experience rapid prepayment caused by the failure of the asset coverage tests. Consequently, the hedge balance grows significantly larger than the reduced balance of the liabilities, and the transaction becomes overhedged. Under stressed economic conditions, which are typically coupled with decreasing interest rates, a CDO can suffer from both an asset/liability balance mismatch and an increase in the periodic payments to the hedge counterparty. This phenomenon has plagued high-yield CBOs over the past few years.

Floating-rate bank loans eliminate the asset/liability mismatch and greatly reduce the interest rate risk in the transaction. Therefore, CLOs are unlikely to suffer from the double blow of being overhedged in a decreasing interest rate environment under stressful credit conditions.

Price Stability

High expected recoveries and a surging demand for loans (especially from CLO issuers) are two of the primary reasons why loan prices have remained very stable over the last few years. Although CDOs are not NAV-based vehicles, price stability is very important for two reasons: credit risk sales and reinvestment into new assets. Regarding the former, if a CLO manager sees signs of credit deterioration at a

Loan prepayment rates provide CLO managers

with additional flexibility.

CLOs do not require interest rate hedging.

The price stability of leveraged loans is

important for reinvestment and credit

risk sale considerations.



company, he or she may decide to sell the loan. The ability to sell this asset into a market that has very solid price stability can be a powerful mitigant to loss of par in the CDO. By contrast, as Figure 34 illustrates, if the CDO manager sold high-yield bonds of the same issuer the loss of par could be much more pronounced.

Reinvestment of principal proceeds is the flip side of this issue. Because the loan market (unlike the high-yield bond market) typically does not trade much above par, when a CLO manager has to redeploy principal proceeds into new assets he or she typically will not pay above par for an asset, or if a premium is paid it will be slight as compared with the high-yield bond market.

Figure 34. High-Yield Bonds and Double-B Rated Institutional Loans — Average Bid Prices, Sep 01–Dec 03

80

85

90

95

100

$105

Sep 01 Dec 01 Mar 02 Jun 02 Sep 02 Dec 02 Mar 03 Jun 03 Sep 03

HY Bonds Loans


CLO Collateral Manager

The final driver of CLO out performance is the least tangible, but perhaps one of the most important: the mindset of the bank loan manager. As we have discussed in the previous section, bank loans are prepayable at par without penalty, and as a result, they do not trade much above par. Loan managers recognize this and the good ones focus zealously on credit risk, because they realize that every point of par lost as the result of trading and/or default is exceedingly difficult to counterbalance through the sale of loans trading at a premium. Not all total return high-yield managers have had the same mindset and some high-yield CBOs have suffered as a result.

CLO managers are used to downward rather than

upward pressure in collateral prices.



Conclusion

The impressive growth that the leveraged loan market has displayed over the past decade has been accompanied by greatly improved liquidity and transparency. The benefits of this asset class, which include stable prices and high recovery and prepayments rates, can be accessed efficiently by CLOs. During the previous credit cycle, CLOs on average have demonstrated more stable performance than both high-yield bond CDOs and straight corporate debt. This stability has fueled CLO growth: CLOs now account for almost one-third of the primary CDO market.

The growth has not come without challenges. The recent surge in demand for institutional loans by structured vehicles has resulted in a significant tightening of the loan spreads and increased obligor concentrations among some institutional loan CLOs. CLO market participants have acknowledged this and are now searching for ways to complement their institutional loan CLO portfolios. Alternative loan categories, such as revolving credit obligations, middle-market loans, and European leveraged loans, provide new chances for diversification, yield, and credit stability.



Glossary

Figure 35. CLO Glossary Term Definition Agent One or several lead lenders in a syndicated loan.

Asset-Backed Security (ABS) Bonds or notes backed by loan paper or accounts receivable originated by providers of credit such as banks and credit card companies. Typically, the originator of the loan or accounts receivable paper sells it to a trust, which repackages it as securities. These securities are then underwritten by brokerage firms, which offer them to the public.

Ask Price Also known as the “offer” price, this is the term for the price at which a seller is willing to sell his loan position in the market.

Assignment and Participation Two ways in which trades in the secondary market can be effected. Assignments are usually necessary for the pro rata part of the loan, which can contain a revolving credit facility and therefore has ongoing funding requirements. While participations in facilities with ongoing funding needs are possible, they are less common than assignments.

Assignment Fees A fee paid by an investor to the administrative agent when purchasing a portion of the loan in the secondary market. The fees, usually in the $2,500-$3,500 range, are used to cover the administrative costs associated with transferring the rights to the loan. This fee has remained quite high relative to fees in comparable institutional securities markets, and we expect it to fall as liquidity improves.

Bid Price The price at which a buyer is willing to buy a loan in the market.

Call Protection Established penalties or restrictions on the prepayment of a loan by the borrower for a specified period. Call protection is usually structured into a loan deal at the time of new issue as an incentive for investors to buy the deal. Call protection is generally structured as a premium above par paid by the borrower at the time of prepayment (101, 102, etc.). In some cases, provisions are written into loan facility agreements preventing prepayment by the borrower for a specific period of time. These restrictions on prepayment are called noncallable periods.

Callability Ability of an issuer to call a security, usually for the purposes of refinancing at a lower rate.

Closed-End Fund Funds that trade on stock exchanges, are bought and sold at a price set by the market and, hence, are more subject to market volatility. Because closed-end funds do not need to meet redemptions, they can stay fully invested at all times, and, thus, post higher yields.

Collateral Loans are typically secured by all of the borrower’s assets that are pledged as collateral. There is a risk, however, that the value of the collateral could decline, causing the loan to be partially or completely unsecured.

Collateralized Debt Obligation (CDO) Generic name for an investment-grade bond offering that is backed by high-yielding debt instruments, such as high-yield bonds or leveraged loans. The most common type of CDOs are CBOs (collateralized bond obligations) and CLOs (collateralized loan obligations). In some investment literature the term CDO and CBO are used interchangeably, even though technically CBOs are limited to investments in bonds, while CDOs can invest in other types of debt instruments.

Collateralized Loan Obligation (CLO) A repackaging of leveraged loans by an asset manager who buys the loans in the primary or secondary markets. CLOs are sold to investors in tranches representing different risk and price levels. An arbitrage CLO refers to a CLO in which the asset manager has the ability to buy loan collateral in either the primary or secondary market at attractive spreads and fund this purchase primarily through investment grade debt issued to the international capital markets.

Continuously Offered Fund Funds that are bought and sold at NAV and can be purchased any time. However, these funds can be typically redeemed only once a quarter, and in a few instances, monthly, and usually at a cost.

Credit Agreement The credit agreement is a legal document establishing the terms of the contract between the borrower and the lenders of a loan facility. Credit agreements are drafted and executed at the time a loan is syndicated in the primary market.



Figure 35. CLO Glossary (Continued) Term Definition Crossover Funds Funds that invest in both loans and bonds.

Coupon The coupon on a loan is a measure of interest rate, expressed as the interest premium a loan pays above a benchmark floating interest rate. Typically, loan coupons are expressed as a basis point spread over the Libor benchmark interest rate.

Derived Loan Spread Term for the bond spread after adjustments in JPM's relative value model to make the bond appear as if it had the same properties as a loan (also adjusted bond spread).

Distressed Debt Any debt security trading at or below 80% of par.

Flexed Pricing Widening the spread, a technique used by arranging banks in the primary market to attract investors to participate in syndicated loan deals.

Floating-Rate Paper Paper priced at a fixed spread over a widely accepted base rate (LIBOR or the Prime Rate), and are pegged to the base rate. Floating rate debt is therefore insulated from fluctuations in interest rates.

Hedge Funds Hedge funds are collective investment vehicles, often organized as private partnerships and resident offshore for tax and regulatory purposes. Their legal status places few restrictions on their portfolios and transactions, leaving their managers free to use short sales, derivative securities, and leverage to attempt to raise returns and cushion risk.

Highly Leveraged Loan According to the Loan Pricing Corporation, any loan that is priced at L+225 or greater. Similar to the definition of leveraged loans, this definition can include a great variance of loans due to its reliance on current market conditions.

Institutional Investors Organizations that buy, or are eligible to buy, leveraged loans. Generally, institutional investors buy the (B, C, D, etc.) tranches of syndicated term loans. Institutional investors committed over $27 billion to leveraged loans in 1H 99. The largest institutional investors included prime rate funds and CLO/CDOs.

Institutional Tranches See "Term Loan B, C, D, etc."

Interest Rate Risk Risk inherent in bonds due to their fixed-rate nature.

Letter of Credit Letter of Credit facilities (also called LCs) are essentially liquidity facilities that ensure a borrower’s payment to a business counterparty. Often used to facilitate overseas trade, the LC facility guarantees payment for goods or services ordered by the borrower. Availability for LC facilities is backed by the borrower’s corporate revolving facility and limited the available borrowing capacity of the revolver. In the event that a borrower does not pay a counterparty transaction set up with an LC, the bank will deliver payment and draw down the requisite amount on the borrower’s revolver.

Leveraged Loan According to the Loan Pricing Corporation, any loan that is priced at L+125bp or greater. The credit quality of the loans captured by this definition can vary widely with market conditions as they impact spreads. A comparable definition would include loans issued by companies that have ratings below Baa3/BBB- from Moody's and S&P, and generally a Debt/EBITDA ratio of 4.0 times or greater.

LIBOR Floor An artificial floor for the LIBOR benchmark interest rate applied to an individual loan that establishes a minimum interest rate the loan will pay, regardless of how low LIBOR goes. LIBOR floors are typically structured into loan facilities at new issue as an incentive for investors to buy the deal.

Liquidity The ease with which one can buy or sell securities for cash in a market. Liquidity constraints on leveraged loans have included high assignment transfer fees, a low number of market participants with trading capabilities, and a low volume of trading. We believe that all of these constraints are disappearing, and that liquidity is improving in the leveraged loan market.

Loan Callability Loans can be called at any time; while high-yield bonds typically are noncallable for a set period. Consequently, leveraged loans generally trade closer to par than high-yield bonds.

Loan Syndication and Trading Association (LSTA)

Organization that has furthered the development of the loan market by standardizing settlement documentation and procedures (T+10 settlement dates, commitment letters and assignment agreements).



Figure 35. CLO Glossary (Continued) Term Definition London Interbank Offered Rate (LIBOR) The rate used by banks worldwide as a benchmark for loans made to large commercial and industrial companies. LIBOR

is generally quoted for 30-, 60-, and 90-day periods.

Mandatory Prepayment A repayment of the principal amount of a loan facility by the borrower triggered by an action of the borrower. Mandatory prepayment provisions are written into the credit agreements of loans to require partial or total repayment of the loan facility in certain specific events such as asset sales, debt issuance, equity issuance, material change of control, etc. Mandatory prepayment conditions differ from loan to loan.

Mortgage-Backed Security (MBS) Security backed by mortgages, in which investors receive payments out of the interest and principal on the underlying mortgages.

National Association of Insurance Commissioners (NAIC)

A national organization of insurance regulators involved in the coordination of solvency and market regulatory activities. Solvency regulation seeks to protect consumers against the risk that insurers will fail to meet their financial obligations. The objective of market regulation is to ensure fair and reasonable prices, products and trade practices.

Net Asset Value (NAV) The market value of a fund share. NAV is calculated by most funds by dividing total net assets by the total number of shares outstanding.

Noncall Period The period following the issue of a security during which an issuer cannot call its issue. A 10NC5, for example, is a ten year instrument not callable for five years. Traditionally non call periods have been a standard feature of high-yield bonds, only recently part of a leveraged loan structure.

OID “Original Issue Discount”. A discount provided to investors during the primary market syndication of a new deal. The discount is usually applied on a price basis, so a deal with a 2 point OID would come at a price of 98 or 2 points below par.

Optional Prepayment Optional prepayments of loans are repayments of principal at the choice of the borrower, rather than scheduled in the general amortization of the loan or mandated by the terms of the credit agreement. As optional prepayments are initiated by the borrower, the borrower chooses the timing and amount of the payment. The terms of the credit agreement establish the order in which loan facilities must be repaid (usually the TLA, then the TLB, then revolver outstandings).

Overcollateralized A term describing the security position of an asset-backed facility. An over collateralized asset backed facility has asset coverage in excess of the size of the facility.

Pari Passu A term describing an equal claim by investors on the assets of an issuer of securities. The term is generally used in reference to the seniority of a debt issue with respect to other outstanding debt by the same issuer.

Prime Rate The rate used by banks as a benchmark for loans made to individuals and small businesses. This base rate is not commonly used for corporate loans. The Prime Rate is quoted for an overnight period.

Prime Rate Funds A mutual fund that invests in portions of corporate loans from banks and passes along interest, which seeks to mirror the prime rate, to shareholders. These funds are among the first non-bank investors in the loan market.

Pro Rata The combination of the revolving credit facility and term loan A tranche sold primarily into the bank market. Participating banks receive higher fees and league table titles (Agent, Co-Agent, etc.) by committing above a certain level to the transaction. The tranches are called "pro rata" because banks must commit to equal percentages of the RC and TLA.

Recovery Rates The average rate of recovery of principal for debtholders in the event of a default by the issuer. Generally, recovery rates are higher on debt at more senior levels of the capital structure that have the initial claim on the assets of the issuer.

Recovery Value The market price of a loan immediately following an event of default. Defaulted loans are generally priced based on the market's perception of the likelihood of repayment.

Relative Value A methodology of comparing two different securities to determine which is the best investment (by looking at the comparable yields and risks).

Revolving Credit Facility (RC) An unfunded commitment that can be drawn and repaid at the issuer's discretion. RC facilities are usually of 5 to 7 year maturity, although maturities may be as short as 1 year and as long as 8.5 years.



Figure 35. CLO Glossary (Continued) Term Definition Second Lien A senior secured debt tranche with a secondary claim on an issuer’s assets behind the claim of the first lien debt.

Second lien debt can be either loans or bonds.

Secondary market trading A market where investors and banks can trade their existing holdings of securities. Secondary market trading is a valuable way for leveraged loan holders to diversify their exposure, take advantage of arbitrage opportunities, invest in attractive deals, or reduce exposure to issuers with deteriorating credit fundamentals.

Security The asset collateral of the borrower pledged to repay the lenders in the event of default on a loan facility. Secured loans have a lien against the assets of a borrower, while unsecured loans are generally just guaranteed by a borrower or one of its entities (parent company, subsidiary, joint venture partner, etc.). Secured loans are commonly used in the leveraged loan market.

Seniority The level of a security holder's claim on the assets of the issuer of that security. Generally, senior secured debt holders have the highest claim on the assets of an issuer, while equity holders have the lowest claim.

Sharpe Ratio A measure of risk-adjusted return based on returns and volatility.

Spread The spread on a loan is the amount of yield the loan pays above a benchmark market interest rate given a particular price in the secondary market. Generally, loans are quoted in basis points of spread over the Libor benchmark rate. So for example, a loan that pays 3.0% above Libor would be quoted as L+300 (basis points). Importantly loan spreads are adjusted for price, with the price discount or premium to par being applied to the spread for the life of the loan. Price discounts or premiums have a greater impact on spread the closer a loan is to maturity.

Swap Spread The spread difference between the Treasury and LIBOR curves or the fixed and floating rate markets.

Syndicated Loan A loan provided by a group of lenders (a syndicate group) and is structured by one or several lead lenders (agents). Traditionally, borrowers had a series of bilateral lines with several lenders. Since the mid-1980's, however, the syndicated loan has become the dominant bank credit product for most large companies. Most syndicated loans comprise several facilities or tranches.

Syndicating Agent A league table title for the leading bank or securities firm responsible for the distribution of a new issue of securities to investors, securities firms and other banks.

Synthetic Asset A value that is artificially created by combining other assets, such as securities.

Teflon Market A market in which problems do not stick.

Term Loan A (TLA) Bundled with the RC facility, the TLA generally amortizes through the life of the loan and has the shortest maturity of the term loan tranches. Spread and term are typically identical for the RC and TLA. As part of the pro rata facility of a syndicated loan, TLA tranches are usually held by banks.

Term Loan B, C, D, etc. (TLB, TLC, TLD, etc.)

The institutional term loan tranches of syndicated loans. Payments on these tranches are usually back-loaded, with interest payments constituting the majority of cash flows in early years. Tranches are designated B, C, D etc. on the basis of maturity. Each successive tranche has a maturity later than the previous tranche. For each additional year until maturity, spreads are generally 25 to 75 bps wider.

Total Return Swap A derivative transaction in which party A receives from party B the return on a certain instrument and pays the funding cost (LIBOR + or - a spread). In the event that the instrument swapped yields a negative return at the maturity of the swap, party A would pay party B.

Unsecured A loan in which the lenders do not have a lien against the assets of the borrower. Unsecured loans generally have a guarantee from the borrower or one of its entities. Unsecured loans are commonly used in the investment grade loan market.

Up-Front Fees In a unique feature to loans, investors charge the issuer a fee for investing. These fees vary depending on tranche, size of commitment and the credit rating of the issuer. These fees also offer the investor implicit call protection.

Sources: Barron's Dictionary of Finance and Investment Terms, Portfolio Management Data, "Hedge Fund Dynamics" by the International Monetary Fund, and Citigroup.

FEBRUARY 4, 2004

UNITED STATES

Ratul Roy Glen McDermott

The CDO of ABS HandbookUnlocking Value From a Diverse Asset Class



Su

mm

ary

A High-Growth Market Structured finance securities (SFS), commonly referred to as asset-backed securities (ABS), have continued to grow in popularity with investors worldwide. Assets invested in the SFS market now total approximately US$ 6 trillion versus US$4 trillion for the US corporate bond market. Strong demand from institutional investors has spurred increased use of securitization technology from a variety of borrowers.

Multiple Advantages — Stability, Diversification, Spread Structured finance securities are attractive because they combine portfolio diversification, rating stability, and low average default rates. An added attraction is that they offer a significant spread pick-up over similarly rated corporates.

The Efficiency of a CDO Structure Collateralized debt obligation (CDO) structures are a powerful way to deliver the benefits of this asset class. By combining the relative stability of SFS with the benefits of risk tranching through a CDO, new investment opportunities are created. In this way, the benefits of the asset class are transferred to a variety of investor classes (from first-loss equity investors up to senior note holders) with different risk-return thresholds.

The Specific SFS Collateral Drives Structures Because of the sheer diversity of SFS collateral, numerous structures have evolved. Portfolios of higher-yielding mezzanine securities are most often repackaged through cash flow CDO structures with term-funded notes. Higher- quality, lower-yielding securities require more efficient funding sources (for example, commercial paper or use of a bank’s balance sheet). Most recently, synthetic structures have evolved that provide a bullet maturity note to investors.

Challenges to Growth While ratings on average have been significantly more stable than corporates, contagion is a consideration. Securities that have been downgraded recently have a considerably higher probability of getting downgraded again compared with the entire universe of credits. Other considerations are liquidity constraints in the subordinated tranches and recent relative underperformance within a few niche sectors (for example, manufactured housing).

Recommendations for Investors Investors should seek exposure to the structured finance sector, and we believe tranched structures such as CDOs are attractive vehicles for this purpose, as they provide tailored risk exposure to investors. We suggest that investors use one of the following three strategies or a combination of three strategies – for investors inexperienced with SFS, look for a third-party managed transaction, for investors more concerned with SFS rating stability, invest in a high quality double or triple-A type portfolio with a watchful eye on the more volatile SFS sub-sectors, or finally, use the CDO vehicle to obtain a leveraged exposure on a focused sub-sector with which the investor is already comfortable (for example, real estate).



Overview of the Structured Finance Market

Securitization Fundamentals Structured finance securities (SFS), commonly referred to as asset-backed securities (ABS),6 are the product of securitization technology, first applied to pools of mortgage loans in the 1970s, and have rapidly developed into the biggest component of the global fixed-income markets. Securitization technology relies on the financing of a diversified portfolio of assets by the issuance of structured finance securities collateralized by the cash flow from the pool. Securitization is based on the “true sale” of assets: no creditors of the seller or the originator have any claims on the assets, and the assets should not be treated as part of the estate of the originator if the originator files for bankruptcy.

Basic Structure Figure 36 illustrates the basic structure. The multiple tranches of SFS have differing levels of seniority with the most senior tranches usually rated AAA and the first-loss, residual or “equity” piece being unrated. Losses arising due to default on the pool are generally applied in inverse order of priority (that is, they are allocated first to the residual piece and last to the senior most tranches). The cash flows from the assets are used to pay the interest and principal amounts of the rated tranches in a specified priority or “waterfall”; residual cash is passed through to the first-loss tranche. SFS ratings are a function of the structure and credit quality of the asset pool backing the securitization, not of the seller that is the source of the assets.

Figure 36. Creation of Tranched SFS From a Portfolio of Assets

Originator

Trust

Cash"True Sale"of Assets

Servicer a

SwapCounterparty

LiquidityProvider

First Loss Piece Typically Retained

First LossPiece

Triple-BSubordinated

Double-AMezzanine

Triple-ASenior

Issue Notes

Receive Cash

Investors

a May also be originator. Source: Citigroup.

6 We use the term SFS as a broad umbrella term that covers three main subsectors: residential mortgage backed securities, commercial mortgage backed securities, and asset backed securities. ABS is also used to only cover non real estate SFS.



Roles of Multiple Parties in a Securitization

While sale of the assets is the first step, there are a number of parties involved in a transaction, summarized in the table below. The seller or originator usually retains interest in the transaction by taking a share of the residual piece and it may also service the collateral.

Figure 37. Key Securitization Parties Party Role Originator Sells assets to trust. Motivated by requiring funding and/or lowering regulatory and economic

capital. Usually retains whole or part of first-loss piece, thus retaining interest in performance of transaction.

Servicer Collects and distributes cash from asset pool and deals with defaulted assets. Usually same as originator. Effectiveness of servicer has strong bearing on deal performance.

Issuer Issues SFS and is usually a bankruptcy-remote trust.

Liquidity Provider Advances cash when required to rectify timing mismatch between asset and SFS cash flows.

Swap Counterparty Provides hedges typically to exchange fixed and floating rate-linked cash flows.

Rating Agencies Credit enhancement for SFS, sized to meet rating agency criteria in order to achieve target ratings.

External Credit Enhancer Provides guarantee to SFS if required.

Source: Citigroup.

ABS Market Fundamentals Asset-backed securities’ many advantages have spurred investor demand for this asset class. In parallel, supply has been growing as issuers have used securitization technology to embrace a wider variety of collateral from a broader range of countries. Growth has been in three dimensions — the type of collateral, the type of risk offered (in particular the amount of subordinated risk), and the overall total volume in the United States and Europe.

Collateral Class Growth

There is a wide variety of collateral backing SFSs. The three biggest categories are residential mortgage backed securities (RMBS), commercial mortgage backed securities (CMBS), and asset-backed securities (ABS), whose collateral most commonly comprise consumer receivables (for example, credit cards). There are other types of ABS that are collateralized by more esoteric assets e.g. mutual fund fees. Other types of SFSs comprise CDOs (collateralized debt obligations) backed by bonds, loans and credit derivatives (as well as other structured finance securities), aircraft leases and securitizations of whole businesses (for example, retail chains). A brief description of the various asset classes within the three main categories is shown in the tables below



Figure 38. Major Structured Finance Asset Typesa Residential Mortgage Backed Securities (RMBS)

These are securities that are backed by residential mortgages. The performance of these transactions is affected by the characteristics and quality of the underlying pool of mortgages, and the abilities of the seller/servicer. In addition, prepayment characteristics of the pool influence the expected maturity and price of the bonds. This is a very large component of the SFS market, and there are several types of RMBSs depending on the type of mortgages backing the security.

Agency Mortgages A large proportion of US mortgages are securitized by Government Sponsored Enterprises — GSEs — that is, Ginnie Mae, Fannie Mae, and Freddie Mac. These securities are typically rated AAA, have very low spreads are not covered by the paper.

Residential A Mortgages Residential A Mortgage securities are similarly backed by a diversified pool of first-lien mortgages underwritten to Fannie Mae and Freddie Mac standards.

Home Equity Loans (HEL); Residential B/C Mortgages

HELs comprise the major segment of securitizations of mortgages that do not conform to the GSE criteria. The principal reasons a loan would be pooled in an HEL include larger loan size and/or weaker credit than permissible by the GSEs. Sometimes the first lien mortgages are referred to as Residential B/C Mortgages; while HEL refers only to the second lien.

Commercial Mortgage Backed Securities (CMBS)

CMBS are securities backed by one or more mortgage loan(s) secured on commercial properties. Properties may include office buildings, shopping centers, multifamily housing complexes, industrial buildings, warehouses and hotels. Three subsectors are usually seen: (1) conduit securities, which are backed by a diversified pool of commercial mortgage loans; (2) credit tenant lease securities, which are backed by a pool of commercial mortgage loans on properties leased to corporate tenants; and (3) large loan securities which are backed by a pool of commercial mortgage loans in which the five largest loans in the pool typically comprise more than 20% of the total pool. In the European market, credit tenant leases are often to a single tenant, thus combining credit (single tenant) and property risk.

Real Estate Investment Trusts (REITs) REITS are corporate debt as they are trusts of a company that purchases and manages real estate or real estate loans using money from shareholders. They are viewed similarly to CMBS.

Asset Backed Securities (ABS) Asset-backed securities are collateralized by the cash flows of a variety of financial assets, typically receivables or loans that are originated by banks and other credit providers. These pooled assets can include auto loans, credit card receivables, consumer loans or personal loans and trade receivables. Credit cards and auto loans are the most liquid segment of the ABS market.

Franchise Loans Usually considered as ABS, but there are strong similarities to CMBSs. Franchise loans are secured by the “going concern” value, inclusive of real estate, of each unit within the lender’s pool. Industries included under franchise include quick-service restaurants, car washes, auto dealerships and aftermarket units, and petrol stations.

Credit-Card ABS These are ABS backed by the receivables of credit cards and constitute the largest share of the ABS market. They are normally not encountered in the CDO market because of relatively tight spreads reflecting the large diversity of the pools and liquidity in the credit card market.

Auto-Loan Backed ABS These are backed by a diversified pool of installment sale loans made to finance the acquisition of, or from leases of, automobiles; again, because of low spreads this collateral class is not part of most CDO portfolios.

For more detailed descriptions please refer to publications listed in Appendix C. Source: Citigroup.



Figure 39. Niche Structured Finance Asset Types Manufactured Housing Loans are given to US buyers of manufactured homes, which are dwellings

constructed at a factory and transported in one or more sections to a land site for attachment. Units are financed as personal property when sold without land, or as real estate when land is included. Manufactured housing is a significant form of home ownership in parts of the United States.

12b-1 Fee, Mutual Fund Fee Securitizations

Mutual fund distributors sell the agreed revenue stream from a shareholder in the fund. The stream comprises a separate distribution fee (referred to as 12b-1 fee) and a contingent deferred sales charge to compensate distributor for lost fees in case of early redemption.

Student Loans Loans to students in the United States are of two main types: federally insured student loans with principal insurance of 98% or 100% of the principal amount, ultimately guaranteed by the US Department of Education; and privately insured student loans which only have protection against default via the guaranty of private companies or from reserves pledged to the securitization.

Small Business Loans (SBL) These are backed by the cash flow from general purpose corporate loans made to “small business concerns” (generally within the meaning given to businesses by the United States Small Business Administration).

Stranded Costs As a result of introduction of competition into electric utility industry, the older utility companies were straddled with unrecoverable and sunk costs as a result of previous investments (for example, nuclear and fossil plants). US statutes enabled recovery of these costs through a tariff that is collectable from the utility’s customers; these tariffs are securitized.

Equipment Leases, Aircraft Securities/EETCs

Through leases, users of equipment (lessee) pay owners (lessors) regular payments for use. Collateral types can be relatively small (for example, copiers, PCs, fax machines) with many borrowers or much bigger (for example, aircraft, farming and medical equipment) with fewer borrowers.

Recreational (RV) Vehicle These are loans that are secured by new and used recreational vehicles, automobiles, and light-duty trucks. Lenders are traditional commercial banks and finance companies, as well as some RV manufacturers. RVs are classed as second homes, and loans are tax-deductible.

Government-Sponsored ABS European governments have sponsored ABS where the underlying assets are owned by the government. For example, the Italian government is an active user of securitization to manage its balance sheet with receivables backed by the national lottery scheme or over-due state pension receivables from employers.

Whole Business Securitizations The securitization ring-fences a specialized, typically asset-intensive whole business with strong credit characteristics. What investors rely on are isolated cash flows for repayment. Examples include telecom service providers, healthcare providers, infrastructure-based groups (such as the water utility sector), transport groups and leisure operators.

Other ABSs In addition there are other niche sectors, for example, securitizations from receivables as diverse as timber, tobacco, and timeshare sales. These are a small part of the ABS market.

Source: Citigroup.

Subordinated Tranche Market Growth

While the overall SFS market has grown tremendously over the last few years (we will explore this in the next section on the US and European markets), the strong growth in the subordinated tranche market is just as impressive. This is a powerful indication of investor confidence in the SFS market. Buying centers of these more junior tranches have increased, in particular the structured product vehicles such as CDOs of ABS. These investors are attracted by the yield pick-up in and relative stability of the asset class. The relative lack of liquidity of the mezzanine tranches is not a concern for such buy-and-hold investors, who are attracted by the additional yield. As an illustration, note the decline in percentage of initial ABS ratings that are AAA and AA and the rise in percentage of initial ABS ratings that are A and below.



Figure 40. Annual ABS Ratings at Issuance by Standard and Poor’s, 1985–2002

0

10

20

30

40

50

60

70

80

90

100

1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002

Perc

enta

ge

AAA-AA A and Lower

Source: Standard and Poor’s.

US and European Market Growth

US volume of issuance has increased sharply as illustrated in Figure 41. Also apparent from Figure 41 is the increasing share of nontraditional ABS types. Recent credit card ABS issuance, for instance, represents a much lower share of total issuance than in previous years. Sectors described as “Other ABS” had little presence in the market even five years ago — these non-traditional ABS sectors (such as student loans, 12b-1fees, equipment leases and franchise loans) are becoming a more important market segment.

Figure 41. Annual US ABS and MBS Issuance by Collateral Type, 1996–2002 (US Dollars in Billions)

0

100

200

300

400

500

600

1996 1997 1998 1999 2000 2001 2002

Auto loans Credit cards Man. Housing

Student loans Other ABS Home EquityCMBS

Source: Citigroup.

In contrast to the US market (Figure 41), the European market is not quite as deep. This is evident by comparing the total volume of European ABS issuance in Figure 42 below with the US issuance shown above.7

7 Markets Inc. is a member of the Se does not include agency (GSE) mortgages.



Figure 42. Annual European ABS Issuance, 1999–2003 (Euros in Billions)

0

20

40

60

80

100

120

140

1999 2000 2001 2002 2003

OtherVehicle/ Equipment LeasesMBSCredit CardCMBSAuto Loans

Sources: MCM Structured Finance Watch and Citigroup.

Because the European SFS market is more modest than the US market, it is a greater challenge for investors who wish to get exposure to a diversified portfolio of European SFSs. Access to collateral directly or having an external collateral adviser with access are necessary for success. And because only part of the European issuance is in Euros (see Figure 43), appropriate currency hedges are required for an investor who wishes to make a euro-denominated investment in a diversified portfolio.

For an investor focused on the higher-yielding mezzanine classes of European SFSs, building diversification is constrained by the large share of CDOs and RMBSs. Figure 44 below shows the breakdown of SFS rated single-A and below that were issued since 1999, and also for 2003 alone. About 70% of the supply side of this market consists of RMBSs and CDOs. This issue can be overcome by some CDO structures, described later, that are able to use multi-currency, multi-region portfolios without passing currency risk on to investors.

Figure 43. Currency of Issuance for European SFS, 2003 By Volume, EUR By Tranches

US Dollar

British Pound

Euro

Others

US Dollar

Others

British Pound

Euro




Figure 44. European SFS Issuance in Single-A and Below Rating Classes, 1993–2003 1999–2003 2003

Other 15% Equipment 3%

AutoLoans 4%

CDO 28%

CMBS 7%CreditCard 4%

MBS 39%

MBS 45%

Other 10% Equipment 1%

Auto Loans 5%

CDO 26%

CMBS 7%CreditCard 6%


In the next chapter we explore the features of the SFS asset class, their attractiveness to investors, and some of the challenges to further growth.



Major Characteristics of Structured Finance Securities

Overview The attractions of investing in structured finance securities stem from (1) relative value as compared to other classes of fixed-income securities, (2) structural protections inherent in these securities, and (3) diverse stable asset pools. An investment in an SFS can also provide diversification away from sectors in which an investor has existing exposure (for example, corporate bonds and loans)

In addition to these strong features, we also need to be aware of some of the challenges to growth in this asset class. We address each of these issues in turn.

Relative Value Structured finance securities often offer considerable spread pick-up as compared to comparably rated corporate debt, and it is this spread difference that structured finance CDOs try to capitalize on. This spread differential is primarily due to two key factors: the perceived lower liquidity in the SFS market and the structural complexity of some structured finance securities. Figure 45 illustrates triple-A spreads for different SFS categories, all of which carry higher spreads than similarly rated supranationals and corporates.

Figure 45. Spreads in SFS Market for Triple-A Ratings, as of 30 Jan 04 CMBS 64 Retail Autoa 5 Credit Card 14 Stranded Assets 17 Home Equity 115 Manufactured Housing 165 CDO (leveraged loan) 65a Maturity for all asset types seven to ten years, except for Retail Autos, which is three years. Sources: Citigroup.

Figure 46 shows the spread differential between CMBS and corporate bonds of similar issuers.



Figure 46. Spread Against US Dollar LIBOR of AAA-Rated Corporates and Mortgage-Backed Securities (In Basis Points)

-10

10

30

50

70

90

110

130

150

170

190

13 Jan 95 27 May 96 9 Oct 97 21 Feb 99 5 Jul 00 17 Nov 01 1 Apr 03

CMBS AAA 10-Yr – Corporate AAA 10-YrCMBS A 10-Yr – Corporate A 10-YrCMBS BBB 10-Yr – Corporate BBB 10-Yr

Source: Citigroup.

Figure 47 shows the spreads for home equity loans and compares them with corporates. The case for relative value is even stronger in the current market of tight corporate spreads.

Figure 47. Spread of Home Equity Loans (HEL) and Similarly Rated Corporates, as Measured by the BIG Index, Over Swaps, 13 Apr 00–26 Jan 04

-50

0

50

100

150

200

250

13 Apr 00 13 Jan 01 13 Oct 01 13 Jul 02 13 Apr 03 13 Jan 04

HEL – AAABIG – AAAHEL – ABIG – A

Source: Citigroup.

Structural Protection When analyzing structured finance securities, it is important to consider the following five features that provide protection to noteholders.

True Sale

Securitization is based on the concept of a “true sale”: no creditors of the seller or the originator have any claims on the assets, and the assets should not be considered part of the estate of the originator if it files for bankruptcy.



Asset Over-Collateralization

The asset coverage cushion is calculated by taking the par amount of the tranche in question, plus par amounts of all tranches senior to it, subtracted from the aggregate par amount of assets.

Excess Cash Flow

Positive excess cash flow is generated when cash flows from the collateral assets exceed the amount required to service the rated debt. This also provides a cushion when collateral underperforms, as the cash may be trapped to either pay down liabilities or be deposited in a reserve account to protect against further losses.

External Credit Enhancement

In addition to intrinsic sources of credit protection described above, some transactions rely on a letter of credit from a bank or a monoline insurance guarantee. These are used less frequently today as investors have become comfortable with structured finance technology.

Collateral Servicer

When analyzing SFS, investors and rating agencies also consider the quality and experience of the collateral servicer. The servicer is particularly important for not just cash collection from the portfolio but also for minimizing losses by dealing appropriately with impaired credits (for example, defaulting borrowers on residential mortgages). The originator is often the same entity as the servicer and retains a share of the residual piece.

Collateral Stability SFS have shown lower ratings volatility, on average, than corporates, and we believe this is due to the portfolio diversification and the structural protections that SFS contain.

Diverse Asset Pools

The collateral backing many structured finance securities is comprised of a diversified pool of obligors. For example, a pool backing an RMBS transaction may consist of up to several thousand mortgages. This low borrower concentration protects the structure from idiosyncratic, or event risk.

Rating Stability

The relative stability of SFS as compared with corporates is shown in Figure 48, which summarizes average transition rates. As an example, 1.27% A-rated ABS obligations migrated down to BBB on average each year; the comparable figure for corporates is 5.3%. See Appendix A for the entire transition matrices.



Figure 48. Standard and Poor’s One-Year Rating Transition Activity Within Investment-Grade Categories ABS, 1985–2001

to Corporate, 1983–2001

to

From AAA AA A BBB Sub IG AAA AA A BBB Sub IGAAA 99.79 0.14 0.05 0 0.02 93.18 6.22 0.45 0.09 0.06AA 1.66 96.41 1.73 0.13 0.07 0.62 91.57 7.08 0.53 0.20A 1.18 0.72 96.38 1.27 0.45 0.06 2.21 91.69 5.30 0.73BBB 0.87 1.18 1.34 91.42 5.20 0.04 0.25 4.69 89.26 5.77

Sources: Standard and Poor’s and Citigroup.

Moody’s findings,8 illustrated in Figure 49, show, however, that while below-investment grade (BIG) corporates have had much higher cumulative transition rates to serious credit impairment (rating of Caa and below) than ABSs, the evidence for higher-quality credits contradicts this over longer time periods.

Figure 49. Average Cumulative Ratings Transitions to Caa and Below, 1983–2002

0.0%

0.5%

1.0%

1.5%

2.0%

2.5%

1 2 3 4 5Years

IG

0.0%

5.0%

10.0%

15.0%

20.0%

25.0%

30.0%

35.0%

BIG

SF IG

Corp IG

SF BIG

Corp BIG

Source: Moody’s.

Low Average Default Rates

Structured finance security defaults have been rare. From 1978 to 2002, Standard and Poor’s rated 18,500 structured finance tranches (RMBS, CMBS, and ABS) and during this period only 178 tranches migrated to a D rating, with 64 of these defaults occurring in 2002.9 Moody’s, in their structured finance default study of 13,419 securities rated from 1993 to 2002, identified 326 securities (2.4%) that either had an uncured payment failure or were rated Ca or C.10 Investors should interpret these studies in the context of the following two observations and caveats.11 First, these default studies (as well as the rating transition studies that we discussed in the previous section) report average behavior of a limited data set. There are, in fact, a few niche SFS sectors that have much higher default and transition rates than the average and have had, as a result, a disproportionately negative effect on the

8 Structured Finance Rating Transitions: 1983–2002, Moody’s Investor Services.

9 Broadly speaking, a D rating occurs whenever a structured finance security either misses a timely interest payment or sustains an ultimate loss of principal.

10 Moody’s included Ca and C rated credits in their study because SFSS in these categories are seriously impaired and will ultimately default with the passage of time.

11 Neither study includes collateralized debt obligations.



studies.12 Second, a typical CDO of ABS is not backed by thousands of obligors, so investors should be careful when using average, actuarial methods to analyze a limited pool of obligors. Single names matter in collateralized debt obligations.

Limited Recovery Data

The relatively rare incidence of default in the structured finance market precludes robust recovery rate analysis. The initial results, however, are interesting. Standard and Poor’s tracked the following defaults over the life of its study (as described in the previous section): RMBS (83), CMBS (14), and ABS (17). Recovery rates for these categories were 61%, 66%, and 29%, respectively. Since most of these securities had not reached their legal maturity, the final recoveries will likely be lower by virtue of having further principal writedowns as the securities age13. In the end, the data set is too short to make a strong prediction about future recovery values in SFS.

Challenges While these benefits provide substantial comfort, there are several challenges that the asset class is facing. Key among these are the underperformance within some niche subsectors and the contagion risk of accelerated downgrades within the lower-rated tranches. Other challenges for investors new to the asset class are the long legal maturities, lower liquidity of the mezzanine tranches, and complexity of understanding the large variety of collateral.

Niche Sector Performance Undermines Strong Average Performance

Performance of the SFS sector was mixed in 2002, with some subsectors performing better than others.14 On the positive side, the RMBS and CMBS sectors on average had higher upgrade rates than downgrade rates, with stability tending to be higher at the higher rating categories (see Figure 50).

12 These asset classes include manufactured housing, franchise loans and 12b-1 fees.

13 Market prices could have been another indicator but prices reflect the promised coupon on the defaulted securities as well as the expectation of receiving future principal payments. Coupon payments were not considered in the Standard and Poor’s study.

14 Performance data does not include collateralized debt obligations.



Figure 50. Annual Percentage of Tranches With Rating Category Unchanged or Higher

60

65

70

75

80

85

90

95

100

Aaa Baa Ba

Perc

enta

ge

All SFS ABS CMBS RMBS All Corporates

Source: Moody’s.

Other sectors, such as manufactured housing, franchise loans and 12b-1 fees, fared poorly (see Figure 51). In fact, of the 58 SFS defaults and 395 downgrades that Standard & Poor’s reported, a large portion emanated from the manufactured housing and franchise loan subsectors. As we will discuss later, CDO of ABS investors must be cognizant of this performance divergence by sector and pay careful attention which asset classes a CDO is permitted to invest in.

Contagion Risk

Rating agency studies illustrate that there is a “momentum effect” in SFS rating migration. Securities that have been downgraded recently have a considerably higher probability of being downgraded again compared with the entire universe of credits, and this characteristic is more evident in the SFS market than with corporates. Figure 52 illustrates the point.

Figure 51. Defaults and Downgrades During 2002 Within ABS Transactions Rated by Standard and Poor’s Defaults, Total = 58 Downgrades, Total = 365

Franchise Loans 38%

Manufactured Housing 55%

Auto 2%

Consumer 3%

ABS Other 2% Franchise Loans 15% 12b-1

Fees 12%

Auto 1%

Manufactured Housing 49%

ABSOther 4%

Consumer 19%

Source: Standard and Poor’s.



Figure 52. Downgrade Risk Conditional and Unconditional on Previous Year Downgrade, 1983–2002 ABS Corporates Unconditional 4.15% 14.7%Conditional on Previous Year Downgrade 36.4 25.8

Source: Moody’s.

As Figure 52 illustrates, on average, ABS is more stable than corporates for many of the reasons described in this report. One of the possible causes of contagion is leverage: SFS structures can magnify the impact of a deteriorating asset portfolio. The other unproven possibility is that the degree of correlation within the portfolio has been different from initial expectations. This contagion risk is more prevalent in the junior tranches because of higher leverage as displayed in Figure 53.

Figure 53. Average Rating Change of SFS Downgraded in Prior Year, by Rating Category Rating after Prior Downgrade Upgraded Same DowngradedAa 0.4 83.3 16.4A 0.6 80.9 18.5Baa 1.0 62.3 36.7Ba 16.0 52.8 31.2B 0 44.1 55.6

Source: Moody’s.

How important is the contagion risk to the investor? The main message continues to be that SFSs have proved to be highly stable investments. However, the downward rating change momentum means that even buy-and-hold investors need to pay close attention to their portfolios, and take appropriate trading decisions at any early signs of credit impairment. The growing liquidity of the secondary market, even in the mezzanine classes, has made the risk management more actionable. Further, the ability of a third-party manager to spot and trade out of these situations may well add considerable value in these situations.

Maturity Considerations

Structured finance securities have legal final maturities that can span 30 years or more. The legal final maturity is the last date that any asset in the trust may contractually mature, and this is the period over which agencies stress the transaction to ensure that the rating is valid. Investors normally focus on the expected maturity, which is the date the bond is expected to be fully retired — through amortization, a clean-up call, a scheduled “soft” bullet or an expected refinancing.

The expected life is based on a base-case or expected scenario of collateral prepayments and defaults. This will include collateral prepayment and default. To the extent that these factors behave differently from the initial assumptions, the average life might shorten or lengthen, creating some uncertainty for investors. For example, under expected collateral pool performance (that is, in the absence of large credit losses), mezzanine tranches have lower volatility in average life since the senior tranches are first to receive principal paydowns.

Liquidity Considerations

Secondary-market liquidity in the structured finance market varies depending on the SFS subsector, tranche rating, and size of issuance. In some sectors (for example, US CMBSs) several hundred million dollars worth of triple-A rated certificates can be



bought and sold in minutes at bid/ask spreads of 1bp-2bp. The liquidity in this portion of the SFS market is similar to that of agency debentures and less like that of many other structured credit products.

In other parts of the secondary market for senior SFS tranches, liquidity is less deep, and it declines further in the market for mezzanine and subordinated tranches. Liquidity in the secondary market is less evident when transaction sizes are small, esoteric asset classes are securitized, or in the case of some derivatives-based transactions, where a small slice of risk is sold into the capital markets and the rest is retained by the sponsor. Despite these challenges, liquidity continues to grow in many areas of the subordinate tranche market, including liquidity in CMBSs and CDOs.

Why is secondary-market liquidity important to investors in CDOs of ABS? If the transaction is managed, good secondary market liquidity allows the manager to exit a credit when it starts to deteriorate instead of waiting for it to potentially decline further.

Having examined the strengths of SFS along with some of the challenges, we now turn our attention to the arguments for using CDO technology to unlock the value of this asset class.



CDO of Structured Finance Securities

Investor Motivation CDOs of SFS combine the benefits of structured finance securities with a CDO’s customized structural advantages. As we have explained previously in this report, structured finance securities have good rating stability and low default characteristics on average. These securities have the potential to add diversity and additional yield to any fixed-income portfolio. CDOs are an efficient way of unlocking the economic benefits of owning a pool of SFSs and the form of ownership can be tailored to meet an investor’s requirements. Through tranching, the benefits are transferred to a variety of investor classes with different risk-return thresholds.

A Customized Investment

The tranching of portfolio credit risk allows investors to participate in an asset class at a risk/reward level that is consistent with their objectives. Each tranche has a different leverage ratio, amount of subordination, coupon and rating. As Figure 54 illustrates, after payment of fees, all cash flow is paid in order of priority, from the triple-A rated senior down to the unrated equity. Portfolio losses are applied in inverse order of priority, first to the equity.

For equity investors the ability to obtain non-recourse leverage on a stable asset class is attractive. Although it is the equity tranche that stands to gain most from a strong credit performance by virtue of its leverage, other noteholders also share in the excess spread. First, mezzanine and senior tranches of the CDO have an attractive spread pick-up from corporates and many types of SFSs (for example, credit-card backed ABS). Second, the mezzanine tranches are not only protected by the par subordination from the collateral, but are also able to benefit from diversion of excess spread away from the equity to the rated notes in case of portfolio deterioration.

Figure 54. CDO as Investment Vehicle Allows Risk Participation at Various Levels

Assets Liabilities

Trustee FeesManager FeesHedging Costs

BBB-ratedportfolio of asset-backed securities

"AAA" Notes

"AA" Notes

"BBB" Notes

Preferred Shares

Principal and Interest from

Portfolio Losses on Portfolio

Source: Citigroup.



Relative Value

Real estate-related SFS collateral offers a practical example of why investors like the advantages of a CDO structure. Many of the same investors that buy real estate SFSs (for example, CMBS and REITs) for their portfolio buy the junior tranches of CDOs of real estate collateral. They are attracted by the combination of collateral, non-recourse leverage provided by the collateral and the spread pick-up for the same rating. For example, during the course of 2003, there were times when the spread differential between a BBB-rated tranche of a real estate-focused CDO and a BBB-rated CMBS tranche was close to 200bp. The spread differential at AA-ratings was also significant (for example, AA-rated tranche of a real estate CDO priced at 100bp over LIBOR in August 2003 compared to 45bp for a AA-rated CMBS tranche).

Major Considerations in CDO Investing Like the underlying SFS, the performance of any specific tranche of the CDO of SFS will depend on the performance of the underlying portfolio, the structural protection given to it, and the performance of the collateral servicer, in this case the CDO manager. In addition, during the life of the CDO there are investment guidelines that must be met.

Structural Protections

CDOs contain two broad types of investment guidelines — one group relates to collateral quality and the second group relates to asset versus liability coverage (coverage tests).

Collateral quality tests relate to specific tests that govern average rating quality, diversification within the portfolio, and sizes of individual obligors and sectors. Breaching these guidelines beyond trigger levels leads to restrictions on future trading. These CDO investment guidelines are intended to maintain the portfolio at the minimum standards to which it was rated at closing.

In addition, most arbitrage CDOs contain two types of coverage tests: a principal coverage test and a liquidity coverage test. If these tests are violated, cash is diverted from the normal priority of payments. The first course of action is diversion of excess spread from the equity tranches to the most senior note. Reinvestment of principal ceases, and principal and interest collections are used to accelerate the redemption of the senior notes until these tests are brought back into compliance.15 These triggers function as structural mitigants to credit risk. Because violation of these coverage tests can result in the payment of all cash flow to the senior noteholders (and consequently none to the equity holders), equity and rated noteholders should have a firm understanding of how they function.

15 See ABCs of CDO Equity, G. McDermott, Citigroup, July 2000.



Diversified CDOs Versus Sector-Focused CDOs

CDOs of SFS have referenced a variety of asset types, of which three classes have formed the majority of new issues. They are as follows:

➤ Multi-sector CDOs. These structures aim to build a diversified pool of various SFS collateral types including RMBS, ABS, operating company securitizations and other CDOs. These pools often contain a large amount of obligor and sector diversity that allows the rating agencies to rate the transaction with less required enhancement all other things equal (that is, the structure may issue less equity, resulting in more leveraged returns to the equity). In addition, this diversity partially insulates senior note holders in these CDOs from idiosyncratic risk.

➤ Real estate CDOs. The assets consist mainly of real estate collateral including RMBS, home equity loans, REITs and CMBSs. Real estate market investors typically favor these deals because they afford a leveraged exposure on an asset class that they are familiar with. These pools have less diversity than the multi-sector CDOs and they may carry more correlated systemic risk that in stressed credit environments may be more adverse to note holders, especially senior note holders.

➤ CDO of CDOs. The collateral consists primarily of cash and synthetic CDOs.

As we have discussed previously, certain niche sectors of the SFS market have performed poorly in the last few years, and this has caused many CDO investors to exclude these sectors in newly created multi-sector CDOs. Transactions in 2004 place greater emphasis on the well-tested RMBS, CMBS and well-established ABS sectors. Figure 55 starkly illustrates the sharp shift in the portfolio composition of CDOs of ABSs from 2001 to 2003. While this restriction in portfolio composition may enhance stability in the future performance of SFS CDOs, it may also ironically prevent a manager from finding relative value in certain out-of-fashion sectors. In today’s markets, an astute investor may find attractively priced assets in these out-of-fashion niche sectors (for example, aircraft lease) that enhance portfolio yield and add diversification to the portfolio.

Figure 55. Typical 2001 Vintage SFS CDO Portfolio Versus Typical 2003 Vintage SFS CDO Portfolio

Utility

HEL

Future Flow

Finance

Equipment Leasing

CMBS Conduit

Residential Mortgages

Credit card

Automobile

Banking

Aircraft lease

CDO

Insurance

Intellectual Property

Manufactured Housing

REITS

Timeshare

Small Business Loan

Project Finance

Small Business

Loan

Credit card

HEL

ManuHsng

REITS

Residential A Mortgages

Residential B&C

Mortgages

BankingAircraft lease

CDO CMBS Conduit

Source: Citigroup.



Static Versus Managed Deal

The decision on whether to hire a CDO manager or invest in a static CDO hinges upon the type of collateral in general, and in particular, the investor’s view of each of the names designated for inclusion in the pool. There are two main areas where one could reasonably invest in a static deal. One is a very high quality portfolio (rated AAA/AA) where investors rely on low statistical probabilities of default. In this case, investors need to be mindful of the more volatile SFS subsectors given the leverage in these vehicles. A second strategy is more focused: an astute investor can use the CDO vehicle to obtain a leveraged exposure to an asset class that it understands and within which it can pick credits with confidence (for example, real estate).

Portfolio managers, through the process of initial asset selection and managing the portfolio over time, can also add value to certain SFS CDOs, especially CDOs of mezzanine SFS tranches. Although SFSs show considerable merit as an asset class from both stability and spread perspective, contagion risk, which is more pronounced in mezzanine ABSs, can threaten a transaction. A diligent manager is able to identify to spot and deal with impaired credits, either by working them out or by trading them (liquidity permitting). The manager may also be able to rotate the portfolio among various SFS sub-sectors in light of changing economic fundamentals and market liquidity. This manager can be a traditional portfolio adviser or, in some cases, a focused service provider (for example, a real estate management company). This expertise, however, comes at a cost: management fees which reduces the excess spread within the structure.

Leveraging Stability — Performance of SFS CDOs How have CDOs of SFS performed? For the most part, the stability of the underlying structured finance market has translated to good performance in the SFS CDO market, save for the transactions that were disproportionately weighted in the niche underperforming SFS sectors such as manufactured housing, franchise loans, and 12b-1 fees.

The table in Figure 56 is derived from Moody’s published data on CDO transition rates and shows the one-year weighted average percentage of tranches that have remained at the same rating or have been upgraded.16 One important caveat: CDOs have a relatively short track record as compared to corporate debt and SFSs, and within the CDO category of SFS CDOs have even less history (they were created in 2000).17 It is important to note, this recent history has coincided with a benevolent, low-interest rate environment.18

16 The data has not been adjusted for withdrawn ratings.

17 CDOs of SFS often contain an investment bucket for other CDOs (for example, a synthetic CDO of SFS will usually contain a 15%-20% bucket for synthetic corporate arbitrage CDOs). As can be seen from the figure, recent performance has varied considerably depending on the underlying asset type.

18 The relative brevity of the existence of the CDO of SFSs can be seen from a comparison of the number of years of data available for the various CDO types, and the lower number of data points compared to CLOs and CBOs.



Figure 56. One-Year Weighted-Average Percentage of Unchanged or Upgraded Ratings

CDO of SFS CLOBal. Sheet

US Cash Bal. Sheet

Non-US CashEmerg. Market CBO

Bal. Sheet US Synth.

Bal. Sheet Non-US Synth.

Inv. Grade Arb. Cash

Synth. Arb. Non-US

Synth. Arb. US

Years of Data 3 7 5 3 7 7 5 3 4 1 1Aaa 100.0% 99.3% 96.7% 89.5% 100.0% 89.8% 89.7% 89.5% 95.0% 65.2% 39.1%Aa 98.2 93.5 93.3 78.9 77.8 77.1 86.1 74.0 76.9 41.0 26.5 A 97.9 97.3 94.4 89.3 78.6 73.4 75.5 75.0 70.0 61.0 6.2 Baa 94.7 92.0 94.9 76.5 83.5 77.7 72.5 66.0 80.9 36.8 14.5 Ba 94.7 90.9 93.9 84.6 80.7 70.1 64.6 70.8 100.0 14.6 20.8

Sources: Moody’s and Citigroup.

CDO of SFSs Take Many Forms Because of the sheer diversity of SFS collateral, numerous structures have evolved for the repackaging of SFSs into CDOs. The development of the structures has also been influenced by the requirement of investors, and the availability of any arbitrage opportunities. Fully funded cash flow structures with term-funded notes are best geared to higher-yielding collateral. Synthetic structures, on the other hand, are especially appropriate to handle very high-quality, low-yielding assets.

Mezzanine SFS CDOs

Because of the requirement to place the entire capital structure in the form of term notes, cash flow CDOs require higher-yielding, lower-rated, SFS collateral. Typically the average rating of the collateral is BBB. Also since the assets are physically held by the issuer, any hedging (for example, currency hedging) must be done explicitly within the SPV. As a result of the search for yield especially under current market conditions, these structures have significant buckets in the higher- yielding SFS collateral types (for example, CDOs, sub-prime RMBSs and HELs, whole business securitizations and non-performing loans). Mezzanine SFS CDOs can be structured as multi-sector or sector specific, and an example of each is shown in Figure 57.

Figure 57. Collateral Mix in Diversified Multi-Sector Versus Real Estate CDO Diversified Real Estate

RES B/C 41%

RES A 15%

REITs5%

HEL 13%

CMBS 15%

Cards 4%

CDO5%

Auto 2%

CMBS 64%

REIT 32%

Real Estate CDOs 4%

Source: Citigroup.



The collateral for the real estate CDO contains, as expected, fewer types of SFSs. As a consequence of the lower diversification the real estate CDO requires a greater proportion of subordinated notes as part of the capital structure (for example, a multi-sector CDO would typically require a 6% subordinated note tranche below a BBB-rated tranche compared to 14% for a real estate CDO). Similarly the proportion of AAA rated liabilities are higher for the diversified multi-sector structure (e.g., typically 75% of the liabilities of a multi-sector CDO would be rated AAA, compared to only 65% for a real estate CDO).

Higher-Quality SFS CDO (AAA Through A Rating) Some arrangers have been able to create funding efficiency within the framework of a cash flow CDO by having the most senior part of the capital structure placed in the commercial paper (CP) market, as shown in Figure 58. In a typical transaction, more than 50% of the assets would be rated AAA with the majority of the remainder rated AA. More than 85% of the portfolio would be funded by the issuance of CP and the remainder would be funded by rated term notes and unrated term subordinated notes.

The immediate impact of this is to lower the weighted average liability cost and therefore the minimum average required spread of the collateral to create an attractive equity arbitrage opportunity. Because of the high quality of collateral, the subordinated note tranche may be only 1% of the capital structure and consequently, highly levered. In the structure shown, the funded notes are exposed to the junior risk of the high-grade ABS portfolio, while the senior risk is transferred on an unfunded basis to a credit swap counterparty. The CP structure is popular in the US market where there is appetite for US dollar-denominated CP. The CP structure would be more difficult to execute outside the United States because currency hedging is expensive and would have to be done within the CDO.

Figure 58. How CDOs Funded by Commercial Paper Achieve Funding Efficiency

[A C A

F unding]

[A C A

F unding]

Hi gh

Gr ad eABSMa rk et

Hi gh

Gr ad eABSMa rk et

[A C A

F unding]

Hi gh

Gr ad eABSMa rk et

Manager

InvestmentAdvisor

IssuerHigh Grade

ABSMarket

Cash

ABS

Principal +Interest

Proceeds

A1+/P-1CP

Program

Swap

Cash

Securities

SuperSeniorSwap

Class A

Preferred Shares

Source: Citigroup.



An alternative to the CP route (not shown) is to fund the underlying collateral in the short-term repo market supported by a backstop liquidity facility. Similar to the CP route, the mezzanine and equity risk are transferred through term notes.

The final way of creating funding efficiency is to use the balance sheet of a counterparty to buy the collateral. This is illustrated in Figure 59. A portfolio of higher-quality (rated A and above) SFS are funded on the balance sheet. The portfolio, in this case, is managed by an external portfolio adviser. Mezzanine and senior note holders (Class A through D notes) sell loss protection on the SFS collateral and are paid a spread for the risk. Loss is crystallized through definitions of credit event and loss valuation, discussed in detail later.

The structure also allows the separation of default risk of the portfolio from any market or currency risk (for example, the portfolio may be denominated in several currencies but senior and mezzanine noteholders may only take default risk on the collateral with the amount of any loss being “pegged” to currency exchange rates at the time of transaction closing). The market risks (for example, currency risks) may be transferred to only the equity holder, retained by the arranger or hedged outside the CDO structure. The structure is particularly appealing for investors who are looking for a globally diversified pool of senior mezzanine SFSs motivated by insufficient diversification in their own currency.

Figure 59. Hybrid Structure Using Cash High-Quality Assets but Synthetic Credit Transfer to Noteholders

PortfolioManagement

Manager

Portfolio

Financedon or

off-balancesheet

Issuer

SuperSeniorSwap

Class A

Class B

Class C

Class D

EligibleCollateral

Loss payments

Senior swap premium

LIBOR

PremiumPayment

Losspayments

LossPayments

LIBOR+spread

Cash

Source: Citigroup.

Synthetic High-Quality SFS CDOs

Synthetic CDO technology enables the creation of bullet structures, a clear winner with investors whose guidelines preclude the long legal maturities of most CDOs of SFSs. These sectors can be structured as bullet deals by broker-dealers who buy protection on a portfolio of high quality SFS (typically AA and AAA rated) from



investors up to the CDO maturity date and are prepared to mange the residual tail risk of securities whose maturities extend beyond that date.

Synthetic bullet multi-sector SFS CDOs typically consist of a bucket of synthetic corporate CDOs as it provides a natural fit for the bullet maturity and is also a source of yield enhancement. As in previous structures of high-quality collateral, the search for yield in the current low spread environment means that these structures are not as highly diversified (consisting typically of 50–60 names) and will seek exposure in the higher yielding SFS collateral types (for example, sub-prime RMBSs and HELs, whole business deals, CDOs and nonperforming loans). Investors need to be aware of this since the low statistical default probability of the collateral means that there is much less subordination below the rated notes compared to the traditional cash flow CDOs. Senior investors should also be aware of any significant overlap in the collateral (for example, RMBSs from the same region, or reference corporate names within corporate CDOs in the SFS pool) as the possibility of extreme losses increases with the higher correlation in future default patterns within the pool.

Figure 60 shows a typical structure. The issuing SPV references the junior risk of a diversified SFS portfolio through a credit derivative contract with a swap counterparty linked to the portfolio. The junior risk is transferred to funded term investors who receive a spread over LIBOR on their investment. The proceeds of their notes are invested in eligible collateral; to the extent that there are losses in the reference pool, the eligible collateral is liquidated to make good any losses incurred by the swap counterparty. Loss is crystallized through definitions of credit event and loss valuation. The maturity of the risk is a bullet five- to seven-year period even though some of the SFSs in the reference pool may have longer legal maturities. Based on current market spreads, there is typically little residual spread to pay third party manager fees and as result most transactions are executed on a static basis. Investors can take comfort from the low historical default rate of the collateral.

Figure 60. Synthetic High-Quality CDOs

CreditSwap

Market

Credit Swapson Individual

Reference Entities

SwapCounterparty

Loss Paymentsin Event of Default

PortfolioCredit Swap

Credit SwapPayments

SPV

Collateral

SuperSeniorRisk

Senior AAA

Junior AA Investors

AA Investors

Equity Investors

EligibleInvestments

$

$

Securities

Source: Citigroup.

Investors in synthetic CDOs should be aware of the importance of documentation of credit events and loss valuation. These two are critical because the determination of a credit event should resemble the occurrence of an extreme impairment or default of a



physical asset. Equally, the valuation mechanism should not introduce additional risk for the investor.

The standard corporate credit event definitions do not fully capture the specific nature of SFS. First, events need to be declared with reference to the tranche and not the whole transaction. Second, standard definitions like “failure to pay” need to be modified. This definition would not cover materially impaired coupon-deferrable tranches since these tranches, by definition, can miss coupon. Industry participants have therefore been comfortable with two main definitions — a modified “failure to pay” and a “loss event,” both of which try to capture the commercial reality of an impaired transaction. Importantly, events can be declared, and investors can incur loss, before a security is declared “D.” This is because agencies have obtained comfort that there is very high correlation between the threshold ratings at which these events can be declared (typically double-C) and subsequent defaults on the SFS.

These themes, as well as the loss calculation process once an event is declared, are described in detail in Appendix B.



Investor Strategies

Investors who wish to invest in a CDO of SFSs, but who want to avoid some of the concerns that we have raised in this paper, may choose to adopt one of the following strategies, or a combination of the three:

➤ Focus on specific collateral type. Rather than invest in the underlying collateral, an investor could use the CDO structure to obtain a leveraged return on the same portfolio on either a static or managed basis. The investor may be sufficiently knowledgeable about the asset class to be able to select individual assets and also be comfortable with the tranche position, typically equity and mezzanine, in the capital structure. There is an argument for a relatively undiversified portfolio (for example, a CDO of real estate loans) that may even be static.

➤ Use a third-party manager. An experienced manager can bring a number of benefits, the first of which is careful picking of assets. Through diligent surveillance, a manager may be able to spot problems early and avoid the “contagion” issue. Finally a manager may have easier collateral access by virtue of being a larger investor in SFS through various vehicles, and in a credit deteriorating situation may find it easier to exit a trade.

➤ Invest in high-quality triple-A type investments. An investor may choose to minimize any of the above risks by investing at the senior-most classes of a CDO of SFSs. An alternative that enhances yield is to invest in a junior tranche of a CDO of a diversified pool of the senior most triple-A rated tranches of SFSs. The underlying collateral has a very low probability of default, and the investor leverages this position.



Conclusion

The multi-trillion dollar global structured finance market is characterized by significant sector diversity, rating stability, and low average default rates. Credit impaired assets have been predominately confined to a few niche sectors, while the bulk of RMBS, CMBS and mainline ABS tranches have performed very well through the end of the last credit cycle. Despite this broad, impressive array of choices, investors should remain wary of contagion, the adverse selection of portfolios, and the underperformance of some sectors.

The relative stability of most structured finance asset classes lends itself well to leverage in the CDO context. Numerous CDO structures have been created to extract the value inherent in the structured finance market. The three main categories are (1) CDOs of mezzanine SFSs, (2) cash flow CDOs of high-grade SFSs and (3) synthetic CDOs of high-grade SFSs. Within each of these categories there are a number of different structures that investors can use, depending on their risk tolerance, yield targets, and desired sectors.



Appendix A

Figure 61. Transition Probabilities AAA AA A BBB BB B CCC CC C DCMBSa AAA 99.44% 0.50% 0.06% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00%AA 3.81 94.76 0.98 0.40 0.00 0.05 0.00 0.00 0.00 0.00 A 0.81 3.20 93.54 2.09 0.18 0.12 0.06 0.00 0.00 0.00 BBB 0.23 1.33 1.97 93.68 2.20 0.41 0.18 0.00 0.00 0.00 BB 0.00 0.08 0.47 2.53 94.23 1.18 0.56 0.08 0.00 0.87 B 0.00 0.00 0.00 0.21 2.06 93.51 2.70 0.00 0.00 1.52

RMBSb AAA 99.80% 0.20% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00%AA 7.50 90.40 1.70 0.20 0.10 0.10 0.00 0.00 0.00 0.00 A 3.00 6.10 88.90 1.40 0.30 0.00 0.30 0.00 0.00 0.00 BBB 0.80 3.30 4.10 88.90 1.20 1.20 0.40 0.00 0.00 0.10 BB 0.00 0.00 2.30 7.20 85.90 1.80 1.40 0.40 0.00 1.00 B 0.10 0.00 0.00 0.90 4.70 88.00 3.90 0.40 0.00 2.00

Other ABSc AAA 99.51% 0.39% 0.08% 0.00% 0.00% 0.00% 0.02% 0.00% 0.00% 0.00%AA 1.59 95.45 2.33 0.57 0.00 0.06 0.00 0.00 0.00 0.00 A 1.02 0.82 95.79 1.66 0.27 0.05 0.05 0.02 0.00 0.32 BBB 0.86 0.92 1.28 90.33 3.24 2.63 0.37 0.00 0.00 0.37 BB 0.00 0.81 0.27 2.17 83.74 7.32 3.25 0.81 0.00 1.63 B 0.00 0.00 0.00 0.00 0.60 54.76 29.76 4.17 0.00 10.71

CDOd AAA 96.60% 1.40% 1.20% 0.70% 0.00% 0.10% 0.00% 0.00% 0.00% 0.00%AA 0.30 90.30 3.50 4.10 1.50 0.30 0.00 0.00 0.00 0.00 A 0.00 0.30 89.20 5.50 2.90 1.30 0.80 0.00 0.00 0.00 BBB 0.00 0.00 0.30 88.90 2.70 3.30 4.50 0.30 0.00 0.00 BB 0.00 0.00 0.00 0.40 88.80 1.70 5.40 2.90 0.00 0.80 B 0.00 0.00 0.00 0.00 0.00 84.40 9.60 6.00 0.00 0.00

Corpe AAA 93.05% 6.29% 0.46% 0.15% 0.05% 0.00% 0.00% 0.00% 0.00% 0.00%AA 0.59 91.03 7.57 0.61 0.06 0.11 0.02 0.00 0.00 0.01 A 0.05 2.10 91.49 5.61 0.47 0.19 0.04 0.00 0.00 0.05 BBB 0.03 0.22 4.38 89.14 4.63 0.94 0.27 0.00 0.00 0.39 BB 0.03 0.09 0.43 5.97 83.02 7.75 1.21 0.00 0.00 1.50 a Source: “Rating Transitions 2002: Respectable Credit Performance of US CMBS,” S&P, January 16, 2003. b Source: “Rating Transitions 2002: A Quarter Century of Outstanding Credit Performance of US RMBS,” S&P, February 6, 2003. c Source: “Rating Transitions 2002: US ABS Weather a Turbulent Year,” S&P, January 31, 2003. d Source: “Rating Transitions 2002: Global CDO and Credit Default Swap Rating Performance,” S&P, March 13, 2003. See Appendix C. e Source: “Corporate Defaults Peak in 2002 Amid Record Amounts of Defaults and Declining Credit Quality,” S&P, February 2003.



Appendix B

Credit Event and Loss Valuation in Synthetic SFS CDOs The standard corporate credit event definitions do not fully capture the specific nature of SFSs. A simple example is a synthetic CDO of corporate reference entities. A certain number of reference entities within the portfolio can be affected by a credit event, for example, ISDA’s definition of bankruptcy without having any impact on a senior CDO tranche while eroding the junior CDO tranche. In such a case, any credit event language should be directly linked to the tranche, and not the CDO as a whole.

Another example is inapplicability of a standard “failure to pay” definition on a coupon deferrable tranche. Even if there is little likelihood of coupon ever being reinstated, the event cannot be declared because the tranche is technically allowed to miss coupon. A corporate “bankruptcy” type definition cannot be used because the CDO issuer is constructed to be bankruptcy-remote; the commercial reality, however, is that the CDO does not have sufficient assets to repay the junior note holders. Therefore a definition that captures this commercial situation is required. Typically two main credit events are used in synthetic CDOs of SFSs:

➤ Failure to pay. A credit event is triggered when interest and/or principal on the reference obligation has not been paid when due and payable (with the exception of ‘PIK’ assets rated as or below a specific rating threshold). This implies that a PIK security that is missing its coupon is not affected until its rating has been downgraded below the trigger (typically Ca for Moody’s).

➤ Loss event. A credit event is triggered when there is a principal reduction as a result of the allocation of defaults, chargeoffs or losses to the SFS in accordance with its terms. If this reduction is reversible rating agencies usually ensure that it is mathematically impossible for the writedown to be reversed assuming no further defaults in the collateral pool and no change in the prepayment rate. Often triggering a “Loss Event” also requires separately that the SFS tranche to have migrated below a specified rating threshold to ensure that the writedown is not because of technical reasons that are not related to true credit deterioration.

Therefore, credit events can be legitimately declared even when the SFS is rated above “D.” This is because agencies from their SFS rating transition studies understand that there is very high correlation between the threshold ratings and defaults on the structured finance tranches that were rated as such. However, to ensure that a credit event is legitimate it is common to ask for simultaneous tests to be performed. An example is the requirement that the tranche should have missed at least two consecutive coupons, and Moody’s should rate the note Ca.



After a credit event has been declared, the documentation allows for two different contract settlement methodologies: physical delivery and cash settlement. Within each of these two main methods, numerous variations can, and do, occur.19 The valuation mechanism is important because depending on the method used, one can have a wide dispersion of recovery values for the same reference credit.

Physical delivery in SFS is typically restricted to the specified “reference obligation.” For investors, this offers more protection than the delivery option that the protection buyer has with corporate credit default swaps.

There are differences in cash valuation too. Unlike corporates where final cash valuation depends on market price of a reference credit, the loss amount for a SFS can sometimes be determined as an exact amount lost by the holder of the reference obligation. For example, if a writedown occurs on an ABS tranche, the protection seller would pay the protection buyer for the full amount of the writedown. If more of the tranche were written down at a future point in time, the additional loss amount would be paid at that time. In such a case, there is no market value risk. An exception is a synthetic CDO within a CDO of SFS portfolio where the individual CDO tranches may be written down but the amount of writedown depends on the market-based bids for the reference credits within the CDO.

In those circumstances where a cash bid is required from the market for an affected SFS, the market for distressed securities is very illiquid. There are a number of ways that synthetic CDOs can be structured so as to lead to preferable valuation treatment for investors. A sufficiently long interval between the occurrence of a credit event and valuation helps to avoid the “lowball” bids that are often prevalent immediately following a credit event. The requirement for multiple dealer bids, repeated dealer bids over a period unless a threshold minimum value is achieved, the ability for certain investors to have a last look at the transaction and the option of physical settlement all benefit the synthetic CDO structure. In addition, rating agencies are penalizing structures with very short valuation timescales by reducing the recovery assumptions (in addition to linking the haircut with the illiquidity of the SFS collateral type) — this leads to higher subordination for the rated tranches.

Regrettably the market for synthetic CDOs of SFSs has not been around long enough for us to make any observations about any differences in the occurrences of credit events, or achieved recovery values, between cash SFS and synthetic SFS linked structures. However, after learning from the synthetic corporate CDO market, investors should consider the features that may have an impact on the market-based valuation of any synthetic structure. This should however be seen in context with the many advantages that a synthetic variation of a CDO of SFSs brings to investors.

19 In a recent survey, Fitch found that there was little consistency in the documentation of the valuation methods in synthetic CDOs. This is different from the interbank market where, overwhelmingly, settlement is occurring through physical delivery of a reference asset. BBA estimates 70%–85% of all contracts are physically settled in the interbank market.



Appendix C

Citigroup Research on Specific SFS Asset Types

ABS

A Primer on Autos and Credit Cards, CABS, Mary E. Kane and Peter DiMartino, Citigroup, January 5, 2001.

A Fresh Look at Credit Card Subordinate Classes, Peter DiMartino, Citigroup, August 2002.

Centex Overview — Stable Prepayments, Improving Credit, and SSB’s New Prepayment Model, CABS, Ivan Gjaja and Anne Kari, Citigroup, March 26, 2003.

Credit Card Almanac, Mary E. Kane and Anne Kari, Citigroup, January 15, 2003.

“Credit Performance of RFC’s RAMP-RZ Deals,” Bond Market Roundup: Strategy, Ivan Gjaja, Citigroup, June 21, 2002.

Fixed-Rate Loans in RASC-KS Deals – Collateral Trends, Prepayments, and Defaults, CABS, Ivan Gjaja, Citigroup, September 20, 2002.

Guide to Mutual Fund Fee Securitizations, CABS, Mary E. Kane, Citigroup, April 4, 2003.

Option One – From Foundations to Rafters, CABS, Ivan Gjaja and Anne Kari, Citigroup, January 31, 2003.

Prepayments on RFC High-LTV Loans, CABS, Ivan Gjaja, Citigroup, April 12, 2001.

Reading Into Sallie Mae’s Private Student Loan Program, Mary E. Kane, Citigroup, May 8, 2003.

Recreational Vehicle ABS Prime, Citigroup, Mary E. Kane and Peter DiMartino, Citigroup, November 19, 2001.

Student Loan ABS Primer: Education — The Investment of a Lifetime, CABS, Mary E. Kane, Citigroup, September 24, 2002.

CMBS

A Guide to CMBS IO, Darrell Wheeler and Jeffrey Berenbaum, Citigroup, May 29, 2001.

A Guide to Commercial Mortgage-Backed Securities, Darrell Wheeler, Citigroup, January 2001.

CMBS Default Parameters – Enhanced to Reflect Watch List Criteria, Darrell Wheeler, Jeffrey Berenbaum, and Gilbert Chua, Citigroup, November 17, 2003.

Realistic CMBS Default and Prepayment Parameters, Darrell Wheeler and Jeffrey Berenbaum, Citigroup, May 7, 2003.



“CMBS Subordinates — Prospecting a Fairly Valued Sector,” Bond Market Roundup: Strategy, Darrell Wheeler and Rohit Hemdev, September 26, 2003.

“Interest Shortfalls Continue to Impact the CMBS Market,” Bond Market Roundup: Strategy, Darrell Wheeler, September 12, 2003.

RMBS

2000 Guide to Agency Debt Securities, Michael Schumacher, Citigroup, January 4, 2000.

A Guide to the Alternative-A Sector, Peter DiMartino, Citigroup, October 1998.

Modeling of Fixed-Rate HEL Prepayments, Ivan Gjaja, Citigroup, August 1998

Anatomy of Prepayments: The Salomon Smith Barney Prepayment, Lakhbir Hayre, Citigroup, April 13, 2000.

Guide to Mortgage Backed Securities, Lakhbir Hayre, Citigroup, March 1999.

European Research

CMBS Trophy Office Assets, Darrell Wheeler and Shaker Sundaram, Citigroup, May 29, 2001.

European ABS — Outlook for 2004, Stephanie Ziar, Citigroup, January 28, 2004.

Finding Value in Structures, David Newman & Shaker Sundaram, Citigroup, March 26, 2002.

Prepayment Modeling and Valuation of Dutch Mortgages, Lakhbir Hayre, Citigroup, January 7, 2003.

Review of the London Office Market; CMBS Relative Value, EABO, Stephanie Ziar and Darrell Wheeler, Citigroup, May 8, 2003.

Swedish Residential Mortgage-Backed Securities, Stephanie Ziar, Citigroup, February 13, 2003.

The CMBS Market in the United Kingdom, Darrell Wheeler, Citigroup, May 2000.

The Mortgage Market in the United Kingdom, Lakhbir Hayre, Citigroup, July 13, 2000.

The Portuguese Securitization Market, EABO, Stephanie Ziar, Citigroup, October 8, 2003.

UK RMBS Master Trusts, Shaker Sundaram, Citigroup, November 7, 2001.

Uncertainty in Italian Lease ABS, EABO, Stephanie Ziar, Citigroup, November 12, 2003.

FEBRUARY 2002

Ratul Roy Glen McDermott

Synthetic Arbitrage CDOs A Product Evolves and Improves



Su

mm

ary

Synthetic arbitrage CDOs are CDOs that use derivatives technology

to transfer credit risk inherent in a pool of credits to capital markets

investors, while at the same time exploiting spread mismatch

between assets and liabilities within the transaction. Their

emergence has been fuelled by the explosive growth of the credit

derivatives market and by their appealing features, which include

short bullet maturities, brief asset accumulation periods, flexible

structures and spreads that are attractive relative to cash CDOs

and corporate bonds.

Despite these and many other attractions, the growth of this market has been hampered over the past few years by problems relating mainly to structure and to the quality of the underlying portfolio. Now, experience with this asset class has led to three areas of marked improvements that promise to reinvigorate this asset class:

1 Portfolio asset selection methods have become more transparent and sophisticated and the ability to substitute reference credits has become more prevalent;

2 Structures have evolved to balance the risks and rewards more fairly across the entire CDO capital structure; the borrowing of performance triggers from cash CDO structures has made mezzanine tranche return profiles more stable, and

3 Credit default swap documentation has improved and has become standardized as it relates to the determination of a credit event and the valuation of the reference obligation post-default.

This paper has three parts. First, we discuss the size and rapid growth of the credit derivatives market and, by extension, the synthetic arbitrage CDO product. Second, we describe how the product has evolved in the three major improvement areas, and we highlight unresolved issues of which investors should be aware. Finally, we present a relative value analysis of a contemporary synthetic arbitrage CDO structure, one akin to what an investor might encounter in the market today.



Introduction

Evolution of Synthetic Arbitrage CDOs Synthetic arbitrage collateralized debt obligations (CDOs) are the descendants of the first synthetic CDOs that were brought to market in the late 1990s. Despite many attractive features common to most synthetic arbitrage CDOs, a portion of this market has performed below expectations. Today, this asset class benefits from marked improvements made in response to the challenges it has encountered. This paper discusses the evolution of this asset class, and illustrates the improvements in the context of an actual transaction.

Unlike cash-flow CDO structures, where CDO notes are collateralized by a portfolio of cash assets purchased by the issuer with the issuance proceeds, synthetic CDOs transfer credit risk from the CDO issuer to CDO note investors through credit derivatives. Regardless of the risk transfer technology (whether cash or synthetic), all CDOs can be separated into two categories: balance sheet CDOs and arbitrage CDOs. The distinction between the two relates to the motivation underlying the transaction. Balance sheet CDOs allow owners of assets to transfer risk from their books to capital markets, thereby improving capital ratios and returns on equity. The first synthetic balance sheet CDOs were done by banks, which found the synthetic structures attractive because the risk transfer and resulting drop in regulatory capital requirements could be achieved without the expensive and cumbersome true sale of the assets. In contrast, arbitrage CDOs, whether cash or synthetic, are motivated mainly by the purchaser of the most junior CDO liability tranche (also called “equity”, “income notes”, or “first-loss” piece) wishing to capture the excess spread between the return on the CDO collateral and the cost of risk transfer on the liabilities above the equity.

The emergence of synthetic arbitrage CDOs has been fueled by the explosive growth of the credit derivatives market. Synthetic arbitrage CDOs, in their simplest form, transfer the credit risk inherent in a pool of credit default swaps to capital markets investors. As the credit derivatives market has grown, so has the market’s acceptance of credit derivatives as collateral for synthetic arbitrage CDOs. Historically, these transactions had many features that have made them attractive to investors. Key attributes include:

➤ Relative Value: Compared with investment-grade average cash flow CDO tranches, investors can usually pick up spread when investing in synthetic CDO tranches of the same rating. These wider spreads, which also depend on the structure of the CDO as well as the expertise of any third-party manager, are even more attractive given the substantially shorter maturities of many synthetic CDO tranches. Investors also pick up spread compared with similarly-rated corporate credit risk.

➤ Shorter Bullet Maturities: Synthetic arbitrage CDOs often have much shorter maturity profiles (3–7 years) than cash arbitrage CDOs (12-35 years). This broadens the addressable market of CDO investors because some are restricted from investing in the longer-dated cash CDO paper.

Synthetic structures allow for efficient credit

risk transfer

Synthetic arbitrage CDOs transfer credit risk

inherent in a pool of credit default swaps to

capital markets investors



➤ Efficient Liability Pricing: Compared with the cost of funds for a typical cash flow arbitrage CDO, the super senior swap can be priced at a tight level, which may allow for more income to be paid to both equity and junior mezzanine tranches.

➤ Investment-Grade Portfolios: More efficient capital structures and liability pricing allow for higher-quality, lower-yielding portfolios. Reference pools in synthetic arbitrage CDOs usually have no allocation to non-investment grade credits at closing; this is in contrast to cash-flow CDOs, which have investment-grade “average” rated collateral with a typical allocation of 30% maximum to high yield credits.

➤ Accelerated Asset Accumulation Period: A portfolio of credit default swaps can be assembled very quickly — a reference pool can be accumulated within days compared with weeks or months for cash CDOs. This is especially important in such markets as European investment grade cash, where the liquidity in the corporate bond market does not permit the ramp-up of a diversified portfolio within such a short timeframe.

➤ Flexible Structures: By dynamically hedging credit positions, synthetic arbitrage CDO products have been created where only parts of the liability structure have been placed with investors (for instance, the “single tranche CDO” trades discussed later in this paper). Also, the liabilities may be partially or fully funded depending on the investor’s preference.

➤ Credit Focused: Synthetic CDOs allow investors to make a “pure” play investment in credit because the structures bifurcate the credit component and all other risks associated with a cash asset (such as interest rate risk, currency risk and prepayment risk).

➤ Smooth Execution: Structures are easier to execute and administer than cash flow CDO structures. Legal documentation and other administrative requirements for asset transfer are less burdensome.

➤ Market Access: Credit default swaps can facilitate exposure to entities that are difficult or impossible to acquire in the cash market.

Despite its appealing features, part of the synthetic CDOs market has performed below expectations. The product has responded to the challenge and evolved over the past few years. First, portfolio asset selection methods have become more transparent and sophisticated and the ability to substitute reference credits has become more prevalent. Secondly, structures have evolved to more fairly balance the risks and rewards across the entire CDO capital structure; in particular borrowing features directly from the cash CDO market. Third, credit default swap documentation has improved, accompanied by greater standardization.

In this paper, we discuss the scope and impressive growth of the credit derivatives and synthetic arbitrage CDO markets. We discuss how the synthetic arbitrage CDO product has evolved and improved in the areas of asset selection, structure and credit default swap documentation. We also highlight those areas where additional progress is needed. We conclude the paper with a synthetic arbitrage CDO case study.

Synthetic arbitrage CDOs have evolved and

improved over the last few years



The Credit Derivatives Market

Meeting Demand for Leveraged Credit Risk The demand for credit protection was the biggest initial driver of growth for the credit derivatives market. Subsequently, this has been surpassed by investor demand for leveraged credit risk, which has in turn catalyzed the growth of synthetic arbitrage CDOs. The credit default swap market has grown exponentially since 1997, with little slowdown in sight. As measured by total outstanding notional, the market has grown from US$180 million in 1997 to nearly US$2 trillion at the end of 2002. In its recently published 2001/2002 Credit Derivatives Report, the British Bankers Association (BBA) estimated that the global market would top US$4.8 trillion by the end of 2004.20

Figure 62. Credit Derivatives Market Growth, 1997–2004E (US Dollars in Billions)

$0

$1,000

$2,000

$3,000

$4,000

$5,000

1997 1998 1999 2000 2001 2002 2004

1997/98 Survey

1999/00 Survey

2001/02 Survey

Source: British Bankers Association (BBA) Survey 2002..

As for the typical maturity profile of these instruments, the credit default swaps market generally encompasses investment grade instruments that mature in five or fewer years (see Figures 63 and 64 below). Seventy-seven percent of the outstanding transaction notional is for five years or less with a little less than half of that being equal to five years. The abundance of shorter-dated credit default swaps should at least partially allay investors concerns that the typical five-year synthetic CDO is unable to trade out of names into new names during its life because of the lack of availability of maturities of less than five years.

20 These numbers, however, need to be interpreted with some caution, as has become clear from another recent survey, “Global Credit Derivatives: Risk Management or Risk?” by Fitch Ratings, Special Report, March 2003). On a notional basis, institutions surveyed by Fitch reported gross sold positions in credit derivatives of US$1.3 trillion. The ambiguity around these numbers was evidenced by the range of interpretations that Fitch found in the responses to the survey questions. For example, something as straightforward as gross “sold” exposure was defined differently. For some, it was the notional value of sold credit derivatives. For others, it was the absolute value of sold and bought positions. For a few others, it was based on the net mark-to-market exposure taking into consideration other cash and derivative positions.

At the end of 2002, the credit default swap

market totaled nearly US$2 trillion

The short end of the CDS market is liquid



Figure 63. Credit Swap Maturity Figure 64. Credit Swap Rating

9%

7%

2%

36%

41%

5%

1–3 Months

3–12 Months

1–Below 5 Years

5 Years

Over 5–10 Years

Over 10 Years

22%

67%

11%AAA-AA

A-BBB

BB-B

Source: BBA Survey 2002. Source: BBA Survey 2002.

In terms of liquidity, the market for more actively-traded default swaps is now on a par with the cash market21. Investors in these frequently-traded credits have begun to exploit the relative ease with which they can transact in the default swap market, and this has produced a significant tightening in the default-cash basis. A tightening trend in the default-cash basis is arguably an indication that the arbitrage between the default and cash markets (at least for these liquid names) is all but disappearing and underscores the increased efficiency of the default market. The demand for credit protection from synthetic CDOs is also contributing to spread tightening. The recovery of the credit derivatives market from credit blowups — as measured by the extent to which the market remained two-sided through the period of stress — is additional evidence of market efficiency.

Finally, in terms of transparency, a growing supply of third-party data providers, such as brokers, offers controllers and risk managers an independent source for trade data. A number of key brokers in credit derivatives are rapidly ramping up their data collection efforts and offering the industry a menu of data service options. Dealers are also supporting the release of market-tracking data as a means of generating investor enthusiasm and comfort in the stability of the credit default swap product. There is also evidence of CDO products that reference public market credit indices.

21 “2002 Trends and Outlook for 2003”, Salomon Smith Barney, Credit Derivatives, January 2003

The CDS market is becoming increasingly

efficient, with some CDS names trading tighter than

their cash equivalents



Synthetic Arbitrage CDOs

Market Overview The term “synthetic arbitrage CDO” is an umbrella term covering any product that references credit default swaps (CDS) as its primary source of credit exposure. It is important to note, however, that whereas the CDO issuer obtains risk exposure on a portfolio of credits through a credit derivative, this may not be the ultimate source of risk. Nor is the application of the technology restricted to corporate credit default swaps; it is being applied to ABS, aircraft loans, small company loans and other asset types. In particular, synthetic arbitrage CDOs that reference ABS, rather than corporate credits, obtain that exposure from a counterparty, which would typically carry the portfolio of cash ABS on its balance sheet and buy protection from the CDO through a credit default swap. This paper will focus exclusively on corporate synthetic arbitrage CDOs. Hybrid structures are also possible where the CDO obtains credit exposure in the form of a combination of cash assets and credit derivatives.

Synthetic CDOs and portfolio default swap products have become big end-users of credit default swaps and, as a result, the arbitrage synthetic CDO market has grown in parallel with the credit default swap market. In fact, 22% of the total volume of credit derivatives market in 2002 consisted of synthetic CDOs and portfolio default swaps. Anecdotal evidence suggests that the share is likely to be even higher since the BBA survey was completed.

Figure 65. Breakdown of Credit Derivative Products by Outstanding Notional

Credit SpreadOptions 5%

Asset Swaps 7%

Total ReturnSwaps 7%

CLNs 8%

Basket Product 6%

Portfolio/CLOs22%

Single-Name CDS45%

Source: BBA Survey 2002.

Its growing share of the CDO market in the United States and Europe illustrates the increasing importance of the synthetic arbitrage CDO product. In Europe, the total number of CDO transactions rated by Moody’s increased from 133 in 2001 to 186 in 2002, but that entire increase was due to synthetic CDOs. Of the total of 159 synthetic transactions, 131 were arbitrage transactions. Similarly, total US CDO issuance increased from 135 to 151 between 2001 and 2002, and at the same time synthetic arbitrage transactions, as a percentage of all CDOs rated in the United States increased from 11% to 27%.

In 2002, 22% of the total volume of credit derivatives were

synthetic CDOs and portfolio default swaps

Synthetic arbitrage CDOs are a rapidly

growing segment of the global CDO market



Product Overview In contrast to synthetic balance sheet CDOs, synthetic arbitrage CDOs create leveraged exposure to a portfolio of credits and exploit the yield mismatch between the spreads on these credits and the cost of risk transfer on the synthetic CDO liabilities. The underlying portfolio in a synthetic arbitrage CDO may be static or managed (discussed below) and it is usually comprised of high-grade credits. Likewise, transactions can be full-fledged, multi-tranche, pass-through transactions or bespoke, “single-tranche” structures described later. In the first version, the entire risk of a portfolio of credit swaps is distributed in the capital markets. The second is a customized single tranche structure, usually sold to a single investor, which is made possible by the hedging allowed by credit derivative technology.

Whatever the version of the CDO, a portfolio of default swaps referencing high-grade credits can generate sufficient portfolio income to make the arbitrage opportunity between the assets and liabilities attractive, particularly in view of the frequent spread differential between swaps and cash bonds. Further, since the risk is acquired by the CDO in an unfunded format, a vast majority of the senior risk can also be placed in an unfunded format at tight liability spreads. It is this feature that creates the biggest efficiency in liability pricing in synthetic CDOs compared with cash CDO structures. A cash CDO is able to place its senior liabilities only in a funded format at wider spreads necessitating the inclusion of a higher yielding non-investment grade bucket (typically up to 30%) within the CDO credit portfolio.

Static Synthetic Arbitrage CDOs

Static synthetic arbitrage CDOs occur when an arranger selects a diversified reference portfolio (by obligor, credit rating and industry) and obtains risk exposure on this portfolio of credits by entering into a series of single-name default swaps. In other words, the dealer sells default protection in the market and collects the premium from the counter-parties with whom it has executed default swaps. These default swaps are the assets of the structure (Figure 66).

The dealer then hedges the resulting credit exposure by buying protection from investors that access various layers of credit risk in the reference portfolio. The credit risk associated with the reference portfolio is tranched into multiple layers or classes including an equity (or first loss) tranche, junior and senior mezzanine tranches, a senior tranche, and a super-senior tranche. These tranches, or liabilities, may be partially funded or unfunded. The super-senior tranche usually comes in the form of a credit swap with an OECD (or highly-rated) financial institution, and senior and mezzanine debt tranches often are publicly placed through a special purpose vehicle (SPV). The residual or first-loss equity tranche may be retained by the arranger or placed externally.

Synthetic arbitrage CDOs exploit the

mismatch between the spreads on a pool of

CDS and the cost of risk transfer on the synthetic

CDO liabilities

A dealer sells protection on a diversified pool

of CDS . . .

. . . and it hedges this risk by buying protection

in different tranches



Figure 66. A Static Synthetic Arbitrage CDO

CASH

CREDIT PORTFOLIO

Reference Entity 1

Reference Entity 50

PREMIUM

PROTECTION

SWAPCOUNTERPARTY

PREMIUM

PROTECTION

SPECIAL PURPOSEVEHICLE

Eligible Collateral

PFANDBRIEFE[Aaa/AAA]

PROTECTION

INTEREST

CASH

Super Senior Swap[Aaa/AAA]

CAPITAL MARKETS

Credit Linked Notes[Aaa/AAA]

Credit Linked Notes[A3/A-]

EQUITY

Source: Citigroup.

If an investor requires funded notes, an SPV is established and on one side it acts as counter-party to the portfolio default swap (referencing a certain level of losses in the reference portfolio), and on the other side it issues the notes. The note proceeds are used to purchase high quality securities that will collateralize the portfolio default swap. The collateral securities are highly-rated and they include, but are not limited to, highly-rated sovereign bonds, Pfandbriefe, financial institution bonds, GICs, funding agreements, money market funds, certificates of deposit and repurchase agreements. In the case of a credit event on a reference credit, the CDO will make loss payments to the swap counter-party of an amount equal to the loss on the notional amount of the defaulted reference credit. In order to make such payment, the CDO will sell eligible collateral and use the proceeds to settle the credit swap. The losses are allocated to the investors in inverse order of priority, starting with the most junior and moving up to the super-senior tranche.

Unlike the pass-through transaction described above, a “single tranche” CDO bespoke transaction is usually structured around a single investor. In addition to the ease of distribution, and hence accelerated time for execution, single tranche CDOs provide a greater degree of control to the investor. Further, no management fees need to be paid to a separate third-party manager, although the flexibility of the product allows this alternative as well. The transaction is structured around the investor’s appetite for the minimum threshold, tranche size and portfolio characteristics. The CDO can be executed either as a (i) funded note backed by eligible collateral and a credit swap contract between the note issuer and a credit swap counter-party or (ii) as a bilateral contract between two counter-parties, one of which is the protection buyer and the other the protection seller.

Figure 67 shows the funded version. In this example, the class C notes are the only notes issued and the transaction is structured around them.22 A swap counter-party pays a premium on the outstanding notional amount of the class C notes, and this notional amount will not decrease from its initial value unless losses on the reference portfolio exceed a minimum threshold equal to the first loss in the diagram.23 As

22 The other parts of the capital structure may or may not be issued in future.

23 This minimum threshold is also called an “attachment point” or “subordination”.

Issuance can be on a funded or unfunded basis

Bespoke single tranche transactions can be

highly customized based on customer preference



losses on the reference portfolio exceed the threshold level, the outstanding notional is correspondingly reduced to a minimum value of zero. The tranche size (also called tranche notional) dictates the total amount of protection. The sizes of the subordination and of the tranche jointly affect the risk profile of the tranche. Usually single tranche transactions have no waterfall24 and therefore, have simple cash flow profiles. The return is equal to the coupon rate times outstanding notional which is directly reduced by portfolio losses.

Figure 67. Bespoke Single-Tranche CDO

Credit 1

Credit 2

Protection

Premium

Credit SwapCounterparty

Cash

Credit n

Distributable ona 'when-issued'

basis

Super Senior[AAA+]

Class B[AAA]

Class C[AA-]

First LossNR/NR

Interest & Principal a

a Or premium in the case of an unfunded transaction. Source: Citigroup.

The arranger is able to create a single tranche CDO without third-party equity or first loss buyers through the process of dynamic risk management, also known as delta-hedging. The risk of the CDO tranche is impacted by changes in default probabilities of the reference credits, the correlation between the various credits and expected recovery rates. Changes in default probabilities are derived from changes in default swap spreads of credits. Therefore, the credit swap counter-party is able to hedge its position in the single tranche CDO (where it has bought credit swap protection on a portfolio) by a long “protection seller” position in a portfolio of single-name default swaps in the reference credits, with each a notional equal to its “delta” where “delta” is the sensitivity of tranche value to changes in any of the affecting variables, such as single-name credit spreads.

Managed Synthetic Arbitrage CDOs

Both the pass-through and single-tranche transactions described above can be, and increasingly are, structured as managed CDO products. The catalyst for growth of managed structures has been the fact that asset managers have gained experience with trading in the credit default swap market and now wish to grow assets under management. A skilled, experienced manager with a deep research staff and robust risk management systems can mitigate default risk by trading out of deteriorating credits.

24 Cash arbitrage CDOs and pass-through multi-tranche CDOs have a specified priority of payments or “waterfall” valid on each note payment date which allocates income from the collateral to the various tranches and other transaction parties and also diverts cash from junior to senior tranches in the case of breach of certain performance triggers. For single tranche CDOs, there is only a promised coupon on an outstanding notional which may be reduced with defaults on the reference portfolio.

Actively managed synthetic arbitrage CDOs

and lightly managed single tranche CDOs are

two popular types of managed synthetic

arbitrage CDOs



Like many managed transactions, credit substitution rules are generally stringent, especially for deteriorating credits, and strict compliance on the part of the manager to these rules is necessary. The difference between a static transaction illustrated in Figure 66 and such a managed public transaction is that there would be a third-party investment manager and a corresponding investment management agreement between the CDO and the manager setting out the portfolio and trading guidelines, as well as terms of business, including appointment, remuneration and removal.

Figure 68 shows a typical capital structure and pricing associated with a managed synthetic arbitrage CDO.

Figure 68. A Managed Synthetic Arbitrage CDO — Capital Structure and Pricing

Obligations

Rating (Moody’s/ S&P)

Principal (Units

in Euros)

Percentage of Structure (Percent)

Subordination (Percent)

Scheduled Maturity

Date

Final Maturity

Date Expected Yielda

Interest Payment Frequency

Super Senior Not rated 860,000,000 86.0 14.0 Mar 08 Sep 08 [0.1]% Quarterly Tranche A Aaa/AAA 60,000,000 6.0 8.0 Mar 08 Sep 08 3m Euribor + [0.6]% Quarterly Tranche B Aa2/AA 17,500,000 1.75 6.3 Mar 08 Sep 08 3m Euribor + [1.2]% Quarterly Tranche C A2/A 22,500,000 2.25 4.0 Mar 08 Sep 08 3m Euribor + [2.5]% Quarterly Income Notes Not rated 40,000,000 4.0 0.0 Mar 08 Sep 08 [15–20]% Quarterly a Returns are for illustration only and will depend on, among other things, market conditions and transaction type. Source: Citigroup.

Increasingly, the bespoke single tranche CDO is structured in a manner that gives investors the right to substitute credits during the life of the trade. Investors thereby become the mezzanine risk-taker as well as manager — hence the name “managed mezz”. These structures allow, as shown below in Figure 69, for adjustment of the initial threshold depending not just on the default but also on the trading performance of the reference portfolio.

Figure 69. “Managed Mezz” Single-Tranche Product

Credit 1

Credit 2

Protection

Premium

Credit SwapCounterparty

Cash

Credit n

Distributable ona 'when-issued'

basis

Super Senior[AAA+]

Class B[AAA]

Class C[AA-]

First LossNR/NR

Interest & Principal a

Manager

InvestmentDecisions

Source: Citigroup.

Many single tranche CDOs allow investors

substitution rights and are called “managed-mezz”



Product Evolution and Improvement

Portfolio, Structure, and Documentation There has been continuous evolution of the credit derivative asset class and the structure of synthetic CDOs since the first products appeared on the market. The catalyst of this improvement has been the performance of a part of the synthetic arbitrage CDO product, which has been below expectations. The roots of this unsatisfactory performance can be traced to three major areas: (i) portfolio asset selection and management, exacerbated by poor recent performance of the investment grade corporate credit class, (ii) synthetic arbitrage CDOs structures and (iii) credit default swap documentation. Although most of the debate in the marketplace has concerned CDS documentation, we believe the biggest factors in the poor historical performance of many synthetic arbitrage CDOs have been poorly designed structures and poor portfolio selection coupled with an exceptionally adverse credit environment. Recently, the market has addressed many of these shortcomings: structures have improved in terms of more equitable risk sharing, investors are more conscious of the need for diligent portfolio selection, and credit swap documentation has become significantly tighter. In the following, we explore the evolutions and improvements in each of these three key areas as well as highlight areas where we think additional progress can be made.

Asset Selection Techniques Within CDO Portfolios Rated synthetic arbitrage CDOs contain numerous investment guidelines that determine obligor/sector concentration limits, average credit quality, limits on asset maturity and other portfolio characteristics. CDO portfolio managers and dealer trading desks have the freedom to satisfy the guidelines in many different ways (that is, more than one portfolio can be constructed to satisfy any set of CDO investment guidelines). For example, a trading desk may decide (all other things being equal) to select assets with the widest yield that allow it to satisfy all investment guidelines, including the portfolio’s weighted average rating restriction. Alternatively, assuming exactly the same investment guidelines, credits with the tightest spreads can be chosen. These two portfolios, both of which satisfy the same set of guidelines, will perform very differently vis-à-vis the number of defaults, severity of rating transition and total yield.

Using our Citigroup Corporate Bond Database, we recently completed an empirical study that examined asset selection methods, and it concluded that portfolios chosen within the same investment guidelines could have substantial performance variation. In Figure 70 we show the distribution of rating changes over a one-year period for three different portfolio selection schemes, which we term Greedy, Conservative and Random25. As shown, even over a one-year horizon, the credit quality of these three sample portfolios (as measured by the change in the portfolio’s average credit rating) can vary dramatically.

25 The “Greedy” selection method aims to maximize returns by choosing the highest yielding credit satisfying the investment guidelines, while the “Conservative” selection method attempts to control loss by choosing the lowest yielding credits. The “Random” selection scheme is one in which credits are picked at random and satisfy, as in the first two methods, the investment guidelines. Repeating the selection process a number of times using selected initial pools of bonds from our historical database allowed us to create a performance distribution for each of these three selection methods.

Despite the many attractive features of

synthetic arbitrage CDOs, three major

factors have contributed to the

performance of a part of the market being below expectations

The way in which CDO investment guidelines

are satisfied will significantly impact future performance



Figure 70. Distribution of Rating Changes in One-Year Period from Three Portfolio Selection Methods

Distribution of Rating Quality Changes, 1999(BBB to BBB- = -1, BBB to BBB+ = 1)

0%

5%

10%

15%

20%

25%

-1.00 -0.80 -0.60 -0.40 -0.20 0.00 0.20 0.40 0.60 0.80 1.00

Freq

uenc

y

GreedyRandomConservative

Source: Citigroup.

What does this mean for investors? In a nutshell, when the rating agencies assign a new rating to a CDO, they pay scant attention to which particular names are selected. Instead, they place heavy reliance on average, aggregate measures. Thus, depending on which names are selected, additional credit risk may be added to a portfolio that has not been accounted for in the rating process. Ignoring the impact of asset selection during the rating of a transaction may result in a selection bias towards picking riskier names that still satisfy the CDO investment guidelines.

A growing number of market participants recognize these risks and are increasingly turning to such market-based measures as bond/loan prices and debt/equity models such as the Moody’s/KMV Model) to parse the risks (“hot spots”) in credit portfolios.26 Such models give default probabilities based on market-based measures, an example being the EDF scores (expected default frequency) from Moody’s/KMV Model. Figure 71 refers again to our empirical study and it starkly illustrates the impact of using market-based information in selecting the Conservative portfolio based on credits with the lowest KMV scores.

26 Debt/equity models combine each firm’s equity market price (which is forward looking) with its balance sheet information to identify those that are most likely to default.

By themselves, rating agency criteria and CDO

investment guidelines will not prevent “hot spots” in a portfolio.

The introduction of market-based

information in the asset selection process can

dramatically improve the performance of a CDO



Figure 71. Three Portfolio Selection Methods — Number of Defaults Experienced over One Year

Number of Defaults, 1999

0

1

2

3

4

Random Greedy Conservative

Source: Citigroup.

We believe that the increased use of market-based measures (both debt and equity market based) when selecting CDO portfolios will significantly improve the performance of synthetic arbitrage CDOs. In fact, all corporate-backed CDOs stand to benefit: (i) cash, (ii) synthetic, (iii) static, (iv) lightly managed and (v) actively managed. Certain segments of the market have used these tools for some time and we believe broader applicability is imminent in the CDO market.

Structural Evolution of Portfolio Products In our opinion, much of the sub-par performance surrounding the first wave of synthetic CDOs (1998–2001) can be attributed to imprudent portfolio selection in an extremely adverse credit environment coupled with structures that did not equitably share the economic risks and benefits of ownership across the entire capital structure. Many of the early static portfolios were assembled from the credits with the highest yield that satisfied the investment guidelines of the CDO (that is, credits that were trading wide for their rating). As we discuss below, this asset selection strategy is not advisable in the most benign of credit markets and it can spell disaster in a bear credit market.

Investors sold credit protection on these portfolios, and downgrades within the static synthetic CDO portfolio were widespread — many credits in fact defaulted.27 Further, since the junior tranches often did not participate in the upside derived from the excess spread, and there were no triggers to divert cash flow to the structure as the portfolios deteriorated, the negative impact on mezzanine and first loss note buyers was pronounced. Other structural features that impacted performance include:

27 Examples of such fallen angels found frequently in the early trades were Worldcom, Enron, Railtrack and Swissair.

The first wave of synthetic CDOs was

plagued by poor asset selection, deficient

structures and a bear credit market

When portfolios deteriorated, early

structures did not divert excess spread for the

benefit of investors



➤ High Leverage: Early transactions were aggressively structured with thin subordination levels below the mezzanine tranche — some structures had less than 3% equity below a BBB tranche with a 1% single obligor limit or even with a 2% obligor limit.

➤ Static Portfolios: Many of the structures did not allow trading. Therefore, selling an impaired security as a risk mitigating strategy was not an option.

➤ Valuation Procedures: Inflexible settlement mechanics left managers little opportunity to maximize recovery values.

➤ Inexperienced Managers: Experienced third party managers had little experience in the credit default swap market and were unwilling to get involved in what was perceived to be a nascent market.

Despite these setbacks, continued investor demand has fueled several important structural reforms in the synthetic arbitrage CDO market. In fact, synthetic and cash flow CDO structuring technologies are converging, leading to a greater resemblance between cash and synthetic structures and a more equitable risk-reward relationship across the entire CDO capital structure. Simultaneously, investors increasingly appreciate the benefits of active portfolio management and, consequently, many modern structures are either managed by third party money managers or they allow for a certain degree of credit substitution by the key investor (for instance, “managed mezzanine” trades).

Cash Diversion Features

Increasingly, cash flow CDO performance trigger technology is being imported into newly-created synthetic arbitrage CDOs. In the typical arbitrage cash flow CDO structure, these triggers measure the amount of asset and interest coverage, and if these coverage amounts fall below certain levels (triggers), excess spread that would have been paid to the equity is instead used to amortize the most senior outstanding CDO tranche. This continues until these coverage ratios once again exceed their minimum levels. The concept is similar in today’s synthetic arbitrage CDOs, but the remedies are slightly different. In some structures, the triggers will divert cash from the junior classes and from subordinated management fees into a cash reserve account, which is used to defray future defaults and trading losses. In other structures, the triggers restrict the ability of the manager to execute discretionary trades in the case of significant portfolio credit deterioration. These triggers are important as they seek either to stop “leakage” of cash to the equity even as the portfolio is deteriorating or to limit the trading activities of a manager of a poorly performing CDO.

The diversion of cash from the equity tranche inures directly to the benefit of the mezzanine tranche; depending on where triggers are set, it makes the mezzanine return profile more stable. Cash diversion triggers are usually of the same two types as arbitrage cash CDOs — over-collateralization and interest. For synthetic CDO transactions, overcollateralization of the liability tranches is calculated relative to the par amount of the reference pool (some transactions calculate the ratios relative to the total amount of tranches below the super-senior tranche). Figure 72 shows the return profile of two mezzanine tranches, one issued from a structure with an over-collateralization trigger and one without. In both cases, the reference pool characteristics are identical, and the capital structures reflect the presence or

Synthetic and cash flow CDO structuring technologies are

converging, leading to a more equitable risk-

reward relationship across the entire CDO

capital structure

Today’s synthetic arbitrage CDOs often

contain many structural protections that have been borrowed directly from the

cash CDO market

The introduction of interest and principal

coverage tests has made the return

profiles of mezzanine tranches more stable



otherwise of triggers (this results in a lower amount of equity for the structure with the triggers).28 As the figure illustrates, the tranche with the trigger can withstand a significantly higher cumulative default rate while still maintaining a stable return.

Figure 72. Recent Five-Year Synthetic Structure — IRR of Mezzanine Tranche at Various Pool Default Rates Moody’s Pool Diversity29 = 50

Weighted Average Pool Rating Factor 30= 273 Default Timing Pattern = Uniform Scenario Recovery Rate = 40%

Note Coupon = 6.1% fixed 8%

6%

4%

2%

0%

-2%

-4%

-6%

-8%

-10%

IRR

0% 2% 3% 4% 5% 6% 7% 8% 9% 10% 11% 12%1%

OC Trigger

Cumulative Default

No OC Trigger

Source: Citigroup.

In addition to the beneficial impact of performance triggers on the performance of the mezzanine tranche, single obligor concentrations can also strongly influence the return profile of CDO tranches. In the following Figure 73, we show the portfolio loss distribution for an equally weighted portfolio; each asset is rated BBB, has a 10-year maturity and is the same size31. The portfolios differ only in the number of obligors. The key message: the smaller the number of obligors higher is the probability of large losses. For example, in a portfolio with a 0.5% obligor limit the cumulative probability for a loss of 2% or less is 80%, whereas for a 2% obligor limit the equivalent number is 70% (i.e., losses larger than 2% are 10% more likely with the 2% obligor limit portfolio).

28 The tranche in question is immediately above the equity and has a Moody’s A2 rating. The portfolio has a 10-year weighted average rating factor of 273 (Baa1/Baa2), which equates to an average five-year cumulative default rate of 1.8%, according to the most recent corporate default study as compiled by Moody’s.

29 Diversity Score is a Moody’s measure that attempts to reflect a portfolio’s diversification by industry and by obligor. The score is equivalent to the hypothetical number of uncorrelated homogeneous assets in the portfolio. For a fuller description, see “The Binomial Expansion Method Applied to CBO/CLO Analysis”, Moody’s Investors Service, December 1996. See also CDO Month-end Valuations, Citigroup, April 2003.

30 The Weighted Average Rating Factor of the portfolio is weighted by the par amount and Moody’s Rating Factor of each asset. Moody’s Rating Factor is a number corresponding to the rating of each entity and is Moody’s idealized 10-year cumulative default probability (multiplied by 10,000) for that rating.

31 The portfolio loss distribution is derived from “copula” analysis described in greater detail in a later section on Performance Analysis.



Figure 73. Impact of Obligor Concentration on Loss Distribution of Portfolio

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

0.0% 2.0% 4.0% 6.0% 8.0% 10.0% 12.0% 14.0% 16.0%

Cumulative Loss Amount

Prob

abili

ty o

f Los

s (L

ess

than

or e

qual

to)

0.5% Limit1.0% Limit2.0% Limit

Source: Citigroup.

Another significant recent reform is a structural feature that allows the mezzanine tranche to share in some of upside of portfolio performance. A share of the excess spread is trapped in a reserve account and the account balance is allocated to the mezzanine note at the transaction’s maturity in return for a lower rated coupon on the note. This mitigates the situation where the mezzanine note is written down even as the equity is earning current income from excess spread on the portfolio.

In addition to cash diversion features, there are a few other structural trends that bear special mention. More synthetic CDO structures, for example, grant a manager the flexibility to exploit relative value opportunities between the cash and CDS markets. Certain structures have even have a bucket for the CDO to take a short position in a credit. Also, newer structures maximize recovery possibilities by providing flexibility in the timing of cash settlement or allowing physical settlement as an alternative to cash settlement. In addition, the ability to harvest gains and mitigate losses through credit improved and credit risk trading is become more prevalent in all types of synthetic arbitrage CDOs, including the single-tranche variety. Finally, credit support beneath the rated tranches is increasing; transactions with 2-3% equity for a collateral pool with a BBB/A average rating are less in evidence.

In summary, investors must scour deal documents and ask detailed questions concerning synthetic CDO structural mechanics32. The market continues to evolve and protections for investors can still vary depending on the deal. Some of the most significant issues that should be explored are (1) the manager’s trading flexibility within the portfolio and his or her ability to trade out of impaired credits, (2) the manager’s ability to express views on the selection of the deliverable asset and valuation time in case of a credit event, (3) in the case of portfolio underperformance, the structural protections offered by cash diversion triggers to mezzanine tranches, and (4) in light of the maximum single obligor limits in the portfolio, the susceptibility of the tranche to event risk.

32 Differences in rating agency methodologies may result in differences in capital structures for the same portfolio characteristics. A discussion of the various methodologies is outside the scope of the paper. The concern here is for the investor to understand the risks and returns of his investment in light of the transaction characteristics and his or her own position within the capital structure.



Credit Swap Documentation The two main issues surrounding credit swap documentation relate to (i) determination of a credit event and (ii) valuation of the reference obligation post-credit event.

Credit Event Determination

Every credit default swap contract lists credit events that define when a default is deemed to have occurred. Although the International Swaps and Derivatives Association (ISDA) documentation contains six credit events, the corporate credit default swap market generally trades on three events: bankruptcy, failure to pay and restructuring.33 At its most basic level, the continued debate surrounding credit events boils down to breadth of interpretation. As Fitch Ratings commented in a recent report, “ISDA faces the unenviable task of trying to manage conflicts of interest between protection sellers who want the narrowest possible definition of a credit event and . . . protection buyers who need the opposite.”34 If a credit event is interpreted too broadly, CDO investors (as sellers of protection) are concerned that realized losses within the reference portfolio may exceed the historical default statistics upon which the CDO is structured and rated.

With respect to each of the three main credit events, bankruptcy and failure to pay are generally reflected in the rating agency historical default studies and, for the most part, they are fairly straightforward. Recent improvements, nevertheless, have been made with respect to the bankruptcy credit event. The 2003 ISDA credit derivatives definitions clarified and improved this event in two ways. The “action in furtherance of” bankruptcy language was eliminated and a written filing requirement was added in order to provide concrete evidence that companies are unable to pay their debts.

Restructuring, the third main credit event in the corporate credit default swap market, has generated the most debate. This event has been amended and improved numerous times in the past (most recently in February 2003), but buyers and sellers of protection have yet to reach unanimity on a single global standard. At its core, the current definition of restructuring is broad enough to encompass events that may not result in loss on the actual instrument that is referenced by the CDS (a “soft” credit event). Soft credit events may trigger losses in synthetic CDO structures that can be avoided in cash flow CDOs (for instance, Solutia and Xerox). For example, assume a synthetic CDO and a cash flow CDO share a common name and a restructuring credit event has occurred with respect to this name but there has not been a payment default on the underlying referenced bond. The synthetic CDO would be forced to crystallize the loss whereas the cash flow CDO could hold onto the instrument with the potential of being paid back at par at maturity. This discrepancy can be especially pronounced in cash-settled synthetic CDO structures where settlement occurs very soon after a credit event is declared. On a more positive note, early triggering of a credit event under the restructuring clause alone may lead to reduced losses in situations where a credit continues to deteriorate and eventually leads to triggering of other credit event clauses.

33 The remaining three events are obligation acceleration, obligation default and repudiation/moratorium. All are typically used only in the Asian and emerging market credit derivatives markets. See “Single-name Credit Default Swaps“, Salomon Smith Barney, December 2002.

34 “Standardized Documentation for Credit Derivatives Growth”, Fitch Ratings, October 16, 2002.

Protection sellers want credit events to be

narrowly defined, and protection buyers want

the opposite

The restructuring credit event has generated

the most marketplace debate, and it is treated

differently in the US and European markets



In response to this concern, among others, numerous market participants have attempted to tighten the restructuring clause.35 In 2001 in the United States, the Restructuring Supplement (commonly referred to as “modified restructuring”) modified the original ISDA definitions. The supplement (i) limited the ability of dealers to deliver the “cheapest-to-deliver” obligation in situations where restructuring was the only credit event exercised, (ii) required that the referenced obligation be fully transferable, (iii) limited restructuring to multiple-holder obligations in order to avoid moral hazard issues within the bilateral loan market and (iv) clarified that only contractual (not structural) subordination would trigger a restructuring credit event. Importantly, however, all soft credit events were not eliminated. Historically, European dealers have not accepted modified restructuring, resulting in a two-tiered market in which European names traded under the original 1999 definitions and the US names traded under modified restructuring (2001) with arguably some adverse impact on liquidity.

There have been further recent developments. In February 2003, ISDA published its new credit derivatives definitions and 2002 master agreement. The new definitions provide for four alternatives: no restructuring, original restructuring, modified restructuring and “modified-modified restructuring”. “Modified-modified restructuring” was developed by European banks and dealers and is similar to the Restructuring Supplement (2001) except (i) with respect to the “cheapest-to-deliver” issue, the maturity limitation is the maximum of five years and the credit default swap maturity for the restructured obligation; and (ii) loans do not have to be freely transferable, so long as they can be transferred subject only to a consent requirement that cannot be unreasonably withheld or delayed.

Hence, in the wake of the 2003 documentation changes it appears that US and European names will continue to trade under different forms of restructuring (and in some cases even without restructuring) and that some soft credit events will remain within scope of the restructuring definitions. What does this mean for synthetic CDO investors? Currently, there is no panacea for the possibility that a soft credit event may occur within a CDS contract, but there are some important factors mitigating risk that investors should consider. First, all three rating agencies employ significantly more conservative analyses (thereby raising the required level of subordination) when they rate transactions that allow restructuring as a credit event. In the case of Moody’s, for instance, use of the original ISDA definition of a restructuring credit event led to an assumption that the probability of default of the reference pool was increased by 22% (versus 10% for the modified definition).36

Second, if a soft credit event occurs under the restructuring definition (that is, it is not actually a default), the market for this credit default swap contract should reflect this fact and, consequently, it should settle at a high recovery value. Finally, all other things being equal, a credit default swap contract that includes restructuring as a credit event will trade more cheaply than a credit default swap contract that does not. Synthetic CDO investors (if one assumes an efficient market) should be compensated

35 In fact, some monoline insurance companies have tried to remove it altogether.

36 Fitch applies penalties for soft credit events, which in its criteria are restructuring and obligation acceleration. The default assumptions for the corporate entities are increased in line with the rating of the entity and the number of soft credit events, for instance, if both credit events are included the increases for “AA- to AAA”, “BBB- to A+” and “Below BBB-” are 10%, 15% and 20%, respectively.

While the 2001 ISDA supplement clarified

certain issues regarding restructuring,

“soft” credit events were not eliminated.

The 2003 ISDA definitions allow for four different types of restructuring

Although all soft credit events have not been eliminated, all rating

agencies require more subordination for those transactions that allow

restructuring as a credit event



for this additional risk because the CDS contracts (with restructuring) will be acquired into the vehicle at a wider spread than CDS contracts that only have bankruptcy and failure to pay as credit events.

These three arguments aside, what does the history of the market tell us? The initial results are good. Although the market spends a considerable amount of time debating the restructuring definition, the vast majority of credit events that are called are clear. In a recent survey, Fitch37 found that of the 112 called credit events, 94.5% were called under bankruptcy or failure to pay, and old and modified restructuring accounted for only 3.3% of the credit events called. Standard and Poor’s38 made a similar observation in a recent report. Of the 30 credit events that occurred in rated US cash and synthetic CDOs, only four (13%) were triggered by restructuring and not by failure to pay or bankruptcy. A third survey, conducted by the BBA 39, also reports very few disputes as to the legitimacy of declared credit events or other contractual terms. This was notwithstanding the fact that there had been a significant increase in the number of credit events due to the heightened number of fallen angels during the past few years.

Valuation of reference obligation after credit event

After a credit event has been declared, the ISDA documentation allows for two different contract settlement methodologies: physical delivery and cash settlement. Within each of these two main methods, numerous variations can, and do, occur.40 The valuation mechanism is important because depending on the method used, one can have a wide dispersion of recovery values for the same reference credit. In the case of physical settlement, the actual security that is being delivered to the protection seller (this is, the CDO) is not known at the inception of the contract except that the security must be pari-passu or senior to the reference obligation and satisfy some other parameters. What the protection buyer produces is likely to be the security that is the easiest or cheapest to deliver.

Theoretically, with respect to two of the three major credit events in the corporate CDS market (failure to pay and bankruptcy) the cheapest-to-deliver option should not cause large recovery value variations for affected obligations with the same seniority in the obligor’s capital structure: in these instances, recovery values for different debt maturities should converge since all become declared payable. Restructuring, however, may be different. As we have discussed earlier, if a soft credit event does transpire under the restructuring credit event, the referenced issuer may not be, in the traditional sense, in “default” on any of its obligations. In this case, debt instruments of the same issuer with differing maturities can diverge significantly in price, with the longer-dated obligations showing the biggest decline. History has shown that obligations from the same issuer, with very different recovery characteristics, can be delivered after a credit event. Swissair is a case in point. The Swissair obligations were issued by a variety of holding companies and operating

37 “Credit Events in Global Synthetic CDOs:2000–2003”, Fitch Ratings, May 2003

38 “Credit Events and Actual Defaults”, Standard and Poor’s Conference Presentation, May 2003

39 “Credit Derivatives Report 2002”, British Bankers’ Association, December 2002

40 In a recent survey, Fitch found that there was little consistency in the documentation of the valuation methods in synthetic CDOs. This is different from the inter-bank market where, overwhelmingly, settlement is occurring through physical delivery of a reference asset. BBA estimates 70-85% of all contracts are physically settled in the inter-bank market.

Historically, the vast majority of credit

events called have been clear

Depending on the valuation mechanism,

the dispersion of recovery values for the

same reference credit can be wide

If a restructuring credit event is triggered, the “cheapest-to-deliver” valuation option can

cause swings in recovery values



subsidiaries and expected recovery levels were different depending on the cash flows securing the individual assets and the presence (or absence) of inter-company loans.

Recently, ISDA has taken steps to address this issue. Modified restructuring (2001) and modified-modified restructuring (2003) include provisions that are intended to mitigate this cheapest-to-deliver risk. Under the former, the maturity of the deliverable obligation can be no later than the earlier of (i) the restructuring date plus 30 months and (ii) the maturity date of the restructured bond or loan. ISDA also says that the maturity of a deliverable obligation cannot be earlier than the credit swap scheduled termination date or later than the credit swap scheduled termination date plus 30 months. Modified-modified restructuring has similar language but the maturity limitation is pushed out to 60 months. A diligent investor can also remove a lot of uncertainty by thoroughly examining the documentation regarding the nature of the deliverable obligation, much in the same way as a cash CDO manager will analyze the recovery characteristics of an asset before deciding whether to purchase it. In particular, non-standardized documentation should be scrutinized. Investors should also inquire about the manager’s experience with credit default swap documentation, and how much flexibility the CDO structure allows vis-à-vis delivery of assets.

Cash settlement is the second major valuation method and in these transactions the valuation is obtained by a dealer poll typically conducted after a threshold period, with bids obtained from a minimum number of dealers. There can be numerous variations on this general framework; in fact, in a recent survey of synthetic CDOs, Fitch Ratings found over 10 permutations. The top three were: (1) the “highest method”, which uses the highest bid from a single valuation round, (2) the “average highest method”, which uses the average of the highest bid from a multiple valuation rounds and (3) the “highest of the highest method”, which uses the highest bid from multiple valuation rounds.41

There are a number of ways that synthetic CDOs can be structured so as to lead to preferable valuation treatment for investors. A sufficiently long interval between the occurrence of a credit event and valuation helps to avoid the “lowball” bids that are often prevalent immediately following a credit event. The requirement for multiple dealer bids, repeated dealer bids over a period unless a threshold minimum value is achieved, the ability for certain investors to have a last look at the transaction and the option of physical settlement all benefit the synthetic CDO structure. In addition, rating agencies are penalizing structures with very short valuation timescales by reducing the recovery assumptions — this leads to higher subordination for the rated tranches.

Given the wide range of valuation methods, it is not surprising that actual experience has been varied. Figure 74 shows the results of a Standard and Poor’s study of well-known credit events that have occurred in rated CDOs. Importantly, this variation is not unique to synthetic CDOs: the data are collected from both cash and synthetic CDOs. Another rating agency report complements the Standard & Poor’s findings. Fitch Ratings found that recoveries in synthetic CDO were not systematically lower than recoveries in cash CDOs, thereby refuting a commonly-held assumption in the market. Approximately one half of the average recoveries obtained by the individual credits were above the cash recoveries, while the other half was below.

41 “Credit Events in Global Synthetic CDOs:2000– 2003”, Fitch Ratings, May 2003.

ISDA has taken steps to mitigate the “cheapest

to deliver” risk

Cash settlement valuation methodologies

are myriad

Synthetic CDOs can be structured so as to lead

to preferable valuation treatment for investors

Historically, recoveries in synthetic CDO have

not been systematically lower than recoveries

in cash CDOs



Figure 74. Recovery Values Credit Value (%)Enron 11–24Pacific Gas and Electric 78Railtrack 70–89Argentina 12–26Crown Cork 41–69Teleglobe 3Global Crossing 33Xerox 64Worldcom 9–12

Sources: Standard and Poor’s.

In summary, when considering an investment in a synthetic CDO that allows for cash settlement upon the occurrence of a credit event, investors should consider (1) the length of time elapsed before the valuation period begins, (2) the minimum number of bids required, (3) the number of bidding rounds and (4) what method is used to arrive at the valuation level (highest method, average highest method).



Analysis of a Modern Synthetic CDO

Case Study: A Recent Transaction As a concrete illustration of the evolution of the modern synthetic arbitrage CDO, we have chosen a recently-issued transaction and will provide a brief relative value analysis of a mezzanine tranche from it. The transaction (which we will call ABC CDO) is a five-year synthetic arbitrage CDO that is actively managed by a large global institution.

Relative value analysis Synthetic CDOs allow investors to make a “pure play” investment in credit because the structures bifurcate the credit risk component of a reference asset and other risks associated with that asset, including interest rate risk, prepayment risk and currency risk. Thus the key parameters that drive the return characteristics of a synthetic CDO tranche are narrower than the drivers of return for cash CDOs. They are:

➤ The absolute level of defaults

➤ The timing of defaults

➤ The recovery rate on defaulted assets

➤ Trading gains and losses, and

➤ Excess spread

The methods by which industry participants analyze leveraged credit portfolios (and, by extension, the expected return on CDO tranches) have become increasingly numerous and sophisticated in the last few years. In particular, the market has seen the growth of better risk and valuation models for the understanding of portfolio credit risk. These models have moved away from the pricing of synthetic CDOs based on historical default data, and now focus on market prices to create replicating portfolios to mimic the expected pay-off of the original credits. Simultaneously, the replicating portfolio enables the structuring and dynamic delta hedging of single tranche CDO structures described earlier. The purpose of this paper is not to describe these methods in depth; rather here we will discuss a few of the simpler methods very briefly so that investors can get an idea as to how the beginnings of a relative value analysis might transpire.

The simplest (and most limited) analytical technique is the constant default rate (CDR) method. A modified version of the CDR method assumes that the constant default rate is periodically interrupted by default rate spikes, which are meant to crudely mirror that default rate volatility that is present in the corporate credit markets. An alternative method derives a distribution of portfolio losses based on the closing parameters, imposes a timing pattern, and analyses expected performance distribution (for instance, Moody’s BET method42 and Standard and Poor’s Loss Distribution). A third method

42 For a fuller description see “The Binomial Expansion Method Applied to CBO/CLO Analysis”, Moody’s Investors Service, December 1996.

Synthetic CDO structures bifurcate the

credit risk component of a reference asset and

other risks associated with that asset

Portfolio analytical methodologies range

from the simple to the sophisticated



generates asset-specific default (and recovery) scenarios from a Monte-Carlo simulator and analyses the resulting performance distribution. Finally, there is a method that analyses credit portfolios and CDO tranche credit risk in a framework that attempts to predict joint distribution of correlated single name defaults. This method is often referred to as the “copula” analysis.

In this paper, as we look at the ABC CDO trade, we will briefly examine the first and last approaches — constant default analysis is still the most popular way that primary market participants analyze portfolio and CDO performance. Although the CDR method is popular, it has serious limitations in that it will not necessarily reflect the granularity of the portfolio, nor will it highlight the impact of credit “hot spots” in the portfolio that may exhibit correlated default behavior.43 Neither will it exhibit correlation patterns in the obligor default timings. That is why we will also briefly describe the copula approach when we look at the mezzanine tranche from ABC CDO.

Constant Default Rate (CDR) Analysis

Figure 75 shows how the mezzanine tranche of ABC CDO performs under increasingly onerous default rate assumptions. This tranche is 2.25% of the capital structure; it has 4% subordination below it, has a Moody’s rating of A2, is immediately above the equity and has a coupon of 6.1%. The portfolio has a 10-year weighted average rating factor of 273 (Baa1/Baa2) for which the most recent Moody’s historical data gives an average five-year cumulative default rate of 1.8%. The results from the three simple scenarios below show that the tranche can withstand several multiples of the historical default rate (approximately five to six times) without diminishing its yield.

Figure 75. Recent Five-Year Synthetic CDO Transaction — Impact of Default Rate, Default Timing and Recovery on Mezzanine Tranche

Even Pattern is Constant Annual Default Rate, 20% of Cumulative in Each Year Front Loaded is 40% of Cumulative in First Year and 15% Annually Thereafter

Stated Coupon of 6.1% Average Default Swap Premium of 125bp

-

40% Recovery, Even Pattern30% Recovery, Even Pattern40% Recovery, Front-Loaded

0% 1% 2% 3% 4% 5% 6% 7% 8% 9% 10% 11% 12%

Cumulative Default Rate

8%

4%

2%

0%

-2%

-4%

-6%

-8%

-10%

6%

IRR

Source: Citigroup.

43 Hence the growing interest in portfolio loss distribution of individual credit-specific defaults (derived either from an analytical distribution or based on Monte Carlo analysis) with the resultant distribution in tranche loss.

The industry is becoming more

advanced in the way it analyzes credit portfolios

The CDR method is a very blunt tool that allows for a quick,

rough assessment of a CDO security



Credit-Specific Correlated Default Analysis (“Copula” Approach)

The copula approach allows credit analysts to examine the term structure of default rates for a portfolio of credit-risky instruments, assuming a correlation structure. This framework for modeling credit portfolio default is described in detail elsewhere44. Correlated random variables are drawn for each credit to denote the survival time for each credit. The output is a construction of a joint distribution of survival times (time-to-default) for the individual credits. Thus, unlike models such as Moody’s BET or Standard and Poor’s CDO Evaluator, derived distributions of cumulative default, the analysis enables the prediction of both default amount and time. Two sets of inputs are key — assumptions of correlation and single name credit curves.

Single name default information can be given by either market prices or historical information. Market prices imply default probability and can be obtained from such sources as default swap spreads and asset swap spreads. Historical information can be drawn from the rating of the obligor and the historical default probability for that rating bucket. Both have advantages and disadvantages. The main problem with pure spread-based default probability is that spreads contain a lot of non-default information, especially a liquidity premium and in many cases will overstate default risk. Historical default data by rating bucket is relatively clean information but, in many cases, obligor ratings lag specific credit information related to the obligor. For the purpose of the following analysis, we have used historical default data derived from the individual obligor ratings.

Correlation parameters within the copula framework can be interpreted as asset return correlations. In our analysis, we have used a one-factor asset return model45. This model assumes that the correlation between each asset in the portfolio is one constant number. For the analysis described below, a correlation assumption of 25% was used. Most practical uses of the analysis go for either this approach or a two-factor model with two constant numbers — inter and intra industrial correlation. Using the “copula” framework for portfolio analysis enables us also to analyze the value of the same ABC CDO tranche described in Figure 75 for a variety of default curves. These default curves are various multiples of historical default rates for each rating category and are used as the proxy credit curve for each credit within the portfolio. The curves are used to determine the survival time for each credit in each simulation — the higher the multiple, the greater the number of credits that would default in that simulation. In Figure 76, we examine the value of the tranche as measured by the present value of the tranche cash flows discounted at Euribor flat46. As in the analysis above, an asset correlation of 25% was assumed for the portfolio.

44 “A Copula Function Approach to Credit Portfolio Modeling”, The Quantitative Credit Analyst, Citigroup, May 2003

45 Each asset return can be modeled as ri = sqrt (ρ) . r

m + sqrt (1 - ρ) . ε

i where rm represents the common factor return and εi is the

idiosyncratic risk associated with each credit asset. All assets are assumed to have the same correlation factor ρ.

46 In reality investors will demand a risk premium to the value derived from the historical default curve to compensate them for the unexpected loss within the portfolio.



Figure 76. Recent Five-Year Synthetic CDO Transaction — Tranche PV, Discounted at Euribor Flat, Based on Various Portfolio Default Rates Using Historical Default-Based Copula Analysis

Tran

che

PV

Multiple of Historical Default Rate

50% 100% 150% 200% 250% 400%300% 350% 450% 500%

0

20

40

60

80

100

120

Source: Citigroup.

One observation that needs to be made is that, unlike Figure 75, increasing default has an impact on tranche value even at very low default rates. The difference stems from the nature of Monte Carlo simulation — if sufficiently large numbers of simulations are run, there will be a finite number of trials that will result in extreme default values and consequently loss in any tranche within the capital structure. The percentage of such trials will be very small for a low assumed multiple of historical default curve. Nonetheless, that percentage will increase as the multiple increases. This distinguishes it from methodologies where specific default curves are assumed.



Conclusion

Synthetic arbitrage CDOs, a rapidly growing portion of the global CDO market, have become important buyers and sellers of protection in the growing credit derivatives market. While the performance of a portion of the first wave of synthetic CDOs was hampered by poor asset selection in a bear credit market, deficient structures and immature non-standardized documentation for credit default swaps, significant improvements have occurred in all these areas. With respect to optimal asset selection in CDO portfolios, the introduction of market-based information in the asset selection process is improving CDO performance. Regarding structure, synthetic and cash flow CDO technologies are converging, leading to a more equitable risk-reward relationship across the entire CDO capital structure. Finally, in 1999, 2001 and 2003, ISDA tightened CDS documentation concerning the scope and breadth of credit events and the valuation of reference obligations after the declaration of a credit event. These reasons, coupled with all the merits of these products, including attractive spreads relative to cash CDOs and corporate bonds, short bullet maturities, brief asset accumulation periods and flexible structures, lead us to believe that investors should pay serious attention to this asset class.

JULY 2000

Glen McDermott

The ABCs of CDO Equity A Primer on Cash Flow Income Notes Collateralized by Pools of Fixed-Income Assets



Su

mm

ary

Cash flow collateralized debt obligations (CDOs) are formed when

asset-backed structuring technology is applied to a pool of

corporate credit exposures. A special purpose vehicle (SPV) issues

notes and uses the proceeds to purchase a pool of fixed-income

assets, which can include high yield bonds, leveraged loans, and

structured finance obligations. Since most of the issued notes are

highly rated, the CDO can raise a majority of its capital cheaply in

the investment-grade market and invest it more profitably in other

markets including the high yield market. The cash flow CDO

income note (or equity), which receives all residual interest cash

flow after payment of fees and rated note holder coupon payments,

is the main beneficiary of this lucrative arbitrage. Traditional CDO

equity buyers have included sophisticated institutional investors

such as banks and insurance companies. We believe, however, that

cash flow CDO income notes represent a compelling investment for

a broader audience that includes pension funds and high net worth

individuals. The purpose of this report is to explain CDO equity

fundamentals to this wider universe of potential buyers.

Healthy Risk-Adjusted Returns

With an experienced CDO manager, the positive spread differential between the yield on the pool of high yield bonds and the lower borrowing costs of the CDO can produce healthy risk-adjusted returns to income note holders in the range of 15%–20%.

Weak Correlation With Other Asset Classes

Cash flow CDO equity returns are driven by the performance of a pool of high yield bonds and bank loans, asset classes that have not shown a strong correlation with other asset classes such as stocks and investment-grade bonds.

Key Drivers of Value

Key quantitative factors in CDO equity returns include the level and timing of defaults and recoveries and the movement of interest rates. In the Return Analysis section of this report, we analyze how these factors can affect CDO equity returns. In addition, we will explore how three important qualitative factors determine relative value among income note opportunities: collateral manager, asset characteristics, and structure.



Su

mm

ary

Role of the Collateral Manager

The experienced CDO manager can outperform the market and generate outsized equity returns. Managers with strong research teams, broad industry contacts, a deep understanding of the intricate CDO investment rules, robust deal flow, and sophisticated credit monitoring systems have a good chance of outperforming the market. CDOs managed by “two guys and a Bloomberg” are at a distinct disadvantage.

Investment Strategies

Since CDO equity returns rely heavily on the experience of the CDO manager and the time during which the CDO assets were purchased, investors should consider two alternative investment strategies. One is to review the forward CDO calendar and select income notes from various CDO issuers, thereby maximizing diversification among different credits and investment styles. The other is to pre-qualify certain “blue-chip” CDO managers and invest in each of their CDOs over time.



Introduction

CDOs Collateralized debt obligations (CDOs) are formed when asset-backed structuring technology is applied to a pool of corporate credit exposures. Total rated issuance of CDOs has boomed in recent years (see Figure 77).

Figure 77. Moody’s Rated Volume, 1988–1999

0

10

20

30

40

50

60

70

80

90

100

1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 19990

20

40

60

80

100

120

140

160

180

Rated Volume (left scale)

Number of Deals (right scale)

Mill

ions


CDO structures can be segmented into three categories:

➤ Cash flow,

➤ Market value, and

➤ Credit derivative.

Cash flow CDOs, which currently are the most prevalent CDO structures, rely upon the cash flow generated from the pool of assets to service the issued debt. This report will focus on cash flow CDO income notes.

A CDO is created when a special purpose vehicle (SPV) is established to acquire a pool of high yield corporate bonds, bank loans, or other debt obligations (see Figure 78). In order to fund the acquisition of the debt obligations, the SPV issues rated and unrated liabilities. Since the majority of the these liabilities are highly rated, the CDO can raise most of its capital cheaply in the investment-grade market and invest it more profitably in other markets including the high yield market.

Since 1996, CDO issuance has boomed.

This report will focus on cash flow CDO

income notes.

CDOs raise capital cheaply in the

investment-grade market and invest it

more profitably in the high yield market.



Figure 78. Typical CDO Structure

COLLATERALMANAGER

TRUSTEE

ISSUER(SPV)

HEDGECOUNTER-

PARTY

CREDITENHANCER

SELLER

HIGH YIELDBOND

PORTFOLIO

INVESTORS

SUBORDINATEDNOTES

INCOME NOTES

SENIOR NOTES

BOND INTERESTAND PRINCIPAL

CDO NOTES

BOND PORTFOLIO

BONDPORTFOLIO

CDONOTE

PROCEEDS

CDO NOTE PROCEEDS

CDO NOTE PROCEEDS

Seller: Sells the bond portfolio to the issuer.Issuer: Issues CDO notes and uses note proceeds to buy bond portfolio.Investors: Purchase CDO Notes.Collateral Manager: Manages bond portfolio.Trustee: Fiduciary duty to protect investors’ security interest in bond portfolio.Hedge Counterparty: Provides interest swap to hedge fixed/floating rate mismatch.Credit Enhancer: Guarantees payment of principal and interest to note holders. Optional.

Source: Salomon Smith Barney.

In a typical cash flow CDO, the rated liabilities are tranched into multiple classes, with the most senior class receiving a triple-A or double-A rating and the most subordinated class above the income note receiving a double-B or single-B. The ratings on the classes are a function of subordination and how cash flow and defaults are allocated among them. Principal and interest cash flow are paid sequentially from the highest-rated class to the lowest, but if the cash flow is insufficient to meet senior costs or certain asset maintenance tests are not met, most or all cash flow is paid to the most senior class.

Cash Flow CDO Income Notes CDO income notes are typically unrated and represent the most subordinated part of the CDO capital structure.47 These notes will receive the residual interest cash flow remaining after payment of fees, rated note holder coupon and the satisfaction of any asset maintenance tests. Factors that will impact the residual interest cash flow include the level and timing of defaults, the level and timing of recoveries and the movement of interest rates. Income notes returns are generated by capturing the spread differential between the yield on the pool of fixed income assets (the majority of which are high yield bonds and bank loans) and the lower borrowing cost of the investment-grade and the noninvestment-grade debt issued by the SPV. This positive spread relationship can produce risk-adjusted returns to income note holders in the range of 15%–20%.

47 Duff & Phelps Credit Rating Co. has recently developed criteria for rating CDO equity: DCR’s Criteria For Rating “Equity” of Cash Flow CDOs, January 2000.

The ratings on CDO classes are a function of subordination and

how cash flow and defaults are allocated among these classes.

CDO income notes are typically unrated and

they represent the most subordinated part of the

CDO capital structure.

CDO income notes can generate risk-adjusted returns in the range of

15%–20%.



Once a cash flow CDO is issued, the collateral manager will manage the portfolio according to the investment guidelines set forth in the bond indenture and within parameters necessary to satisfy the rating agencies. Pursuant to these guidelines, the manager will sell and buy assets and, during the reinvestment period, will reinvest collateral principal cash flows into new bonds. The investment guidelines typically require that the CDO manager maintain a minimum average rating and portfolio diversity such that any trading will have a minimal impact on the senior CDO bondholders.

The primary responsibility of the cash flow CDO collateral manager is to manage the portfolio in a way that minimizes losses to note holders resulting from defaults and discounted sales. To this end, all note holders rely on the manager’s ability to identify and retain creditworthy investments. A manager’s trading decisions can have a substantial impact on the returns paid to income note holders; the initial asset selection and its trading activity throughout the reinvestment period are critical to achieving high returns. A manager with a deep understanding of the underlying credit fundamentals of each of its investments can make informed, credit-based trading decisions, not trading decisions based on price movements.

Cash flow CDO income notes have many favorable characteristics. Among them are:

➤ Healthy returns. The risk-adjusted internal rate of return (IRR) to income note holders can range from 15%–20%. This return rate compares favorably with that of other investment opportunities (see Figure 79). We will explore the volatility of this return in the “Return Analysis” section of this report.

Figure 79. Historical Returns — Various Asset Classes, 1990–2000 Index

10 Year Average (%)

Standard Deviation

Sharpe Ratio

5 Year Average (%)

Standard Deviation

Sharpe Ratio

Fixed Incomea SSB Brady Bond Index 17.52 15.98 0.78 17.52 17.88 0.69SSB HY Market Index 10.94 6.39 0.92 9.71 5.25 0.86Corporate Bond Index 8.27 4.58 0.70 8.15 4.66 0.63Intermediate Term Treasury Index 7.89 5.79 0.49 7.95 5.78 0.47Mortgage Bond Index 7.84 3.17 0.88 7.93 2.87 0.95SSB Broad Investment-Grade (BIG) Index 7.75 3.88 0.69 7.74 3.74 0.68AAA Rated Corporate Bond Index 7.65 3.70 0.70 7.46 3.68 0.61

Equities Nasdaq Composite 24.50 20.62 0.94 40.17 22.82 1.53S&P 500 18.27 13.42 0.98 28.72 14.07 1.67DJIA 16.17 14.26 0.78 24.56 15.65 1.24aFor more information on Salomon Smith Barney indexes, please see April 2000 Performance, Total Rate-of-Return Indexes, Salomon Smith Barney, May 2, 2000. Source: Salomon Smith Barney.

➤ Lack of correlation with other asset classes. Although there is no established index for CDO income note returns, the performance of the traditional cash flow CDO income note is directly linked to the behavior of a pool of US high yield bonds. US high yield bonds have not shown a strong correlation to other asset classes (see Figure 80).

Trading decisions can have a substantial

impact on the returns to income note holders. A good cash flow CDO manager bases trading

decisions on credit fundamentals rather

than price movements.



Figure 80. Correlation of Various Asset Classes, 1990–2000 SSB HY

Mkt Index 3 Month

Treas. BillBrady Bond

Index CorporateTreasury

7-10 Year BIGAAA Rated Corporates

Mortgage Index

Nasdaq Composite DJIA

3-Month Treasury Bill -0.08 SSB Brady Bond Index 0.47 0.11 Corporate Bond Index 0.53 0.10 0.34 Intermediate Term Treasury Index 0.30 0.09 0.19 0.94 SSB Broad Investment-Grade (BIG) Index 0.41 0.14 0.26 0.98 0.98 AAA Rated Corporate Bond Index 0.45 0.11 0.28 0.98 0.97 0.99 Mortgage Bond Index 0.43 0.22 0.29 0.90 0.87 0.94 0.91 Nasdaq Composite 0.51 -0.05 0.43 0.29 0.16 0.21 0.25 0.21 DJIA 0.53 0.00 0.59 0.38 0.25 0.32 0.33 0.35 0.71S&P 500 0.54 0.00 0.56 0.47 0.35 0.40 0.42 0.42 0.82 0.92


➤ Top-tier fund managers. An income note investment allows an investor to gain exposure to an experienced CDO manager and the healthy returns it can generate, with a smaller initial investment than might otherwise be required.

➤ Access to esoteric assets. An income note investment can be an efficient way for an investor to gain exposure to a variety of esoteric asset classes. Certain asset types, such as leveraged loans, mezzanine loans, and project finance loans, are asset classes to which relatively few investors have access.

➤ Cash flow-based returns. Returns on cash flow CDO income notes are driven by the cash flow generated from the assets, not the market value or the price of those assets. This characteristic enables the investor to mitigate market risk and allows the manager to focus on the underlying credit fundamentals of the high yield collateral. The investment is especially attractive when there is a dislocation in the high yield market due to technical, not credit, factors (e.g., in fourth-quarter 1998). This stands in stark contrast to high yield mutual fund returns, which are sensitive to market value fluctuations.

➤ Diversification. A relatively small investment in a cash flow CDO income note can confer substantial diversification benefits. An investor can gain exposure to 50–120 obligors across 15–25 industry sectors.

➤ Structural protections. Income note holders benefit from a variety of structural features present in cash flow CDOs. Chief among them is the ability to remove the portfolio manager and the right to call the deal after the end of the noncall period.

➤ Front-loaded cash flows. Unlike other alternative investments (e.g., private equity) an investment in cash flow CDO income notes will typically generate cash flow within six months of the initial investment.

➤ Transparency. Income note investments are more transparent than many alternative investments. Every month, the trustee reports, among other things, trading activity, obligor names and exposure amounts, industry concentrations, and compliance or noncompliance with liquidity and asset maintenance tests.

➤ Imbedded interest rate hedges. Many CDOs are floating-rate obligations backed by pools of fixed-rate bonds. In order to hedge the mismatch between the fixed-rate assets and the LIBOR-indexed liabilities, most CDOs purchase a combination of interest rate swaps and/or caps. Although these hedges are bought for the benefit of the rated note holders, they also benefit the income notes as the residual interest beneficiary.



➤ Taxation. Non-US investors are not subject to US withholding tax on dividends and gains from sale, exchange or redemption they receive from their investment. Tax-exempt entities are not subject to Unrelated Business Income Tax.

Although CDO equity has many favorable characteristics, prospective note holders should consider the risks associated with ownership. Some risks include:

➤ Subordination of the income notes. The income notes are the most subordinated notes in the CDO capital structure. They receive interest cash flow only after fees and rated coupon interest are paid, and asset and cash flow coverage tests are satisfied. No payment of principal of the income notes is paid until all other notes are retired and, to the extent that any losses are suffered by note holders, such losses are borne first by the income note holders.

Since the income notes are subordinated, prospective investors should consider and assess for themselves, given the manager’s track record, the likely level and timing of defaults, recoveries and interest rate movements. The following section, “Return Analysis” provides numerous examples that will help investors understand how these variables impact income note returns.

➤ Limited Liquidity and Restrictions on Transfer. Currently, potential income note buyers should not rely on a secondary market for CDO income notes. The investment trades on a “best efforts” basis and in a typical transaction, the income notes will be owned by a relatively small number of investors. Also, before selling an income note in the secondary market, the seller must comply with various regulations that restrict the transferability of certain types of securities.

➤ Mandatory Principal Repayment of Senior Notes. If the aggregate asset balance is insufficient to meet the minimum overcollateralization test or the aggregate asset yield is insufficient to meet the minimum interest coverage test, cash flow that would have been distributed to the income notes will be diverted to amortize the most senior notes. If this occurs, the capital structure will de-lever until the test(s) are brought back into compliance. A de-levering structure will have a negative impact on income note returns.

➤ Reinvestment Risk. During the reinvestment period the collateral manager will reinvest principal collections in additional bonds and loans. Depending on market conditions and the CDO’s investment guidelines, the manager may purchase loans and bonds with a lower yield than the initial collateral (i.e., spread compression), resulting in less cash flow for all note holders.

One way to mitigate some of these risks is to bundle the income note with a zero coupon Treasury STRIP (or other security free of credit risk) and create a principal-protected structured note or unit (PPU) (see Figure 81). These units are designed to protect income note holders from the loss of their initial investment while still providing the potential of some yield upside. Moody’s can rate PPUs Aaa, thereby allowing insurance company investors to gain NAIC1 capital treatment.

Some risks associated with owning CDO

income notes can be mitigated by bundling it with a Treasury STRIP.



Figure 81. Principal Protected Units

$10,000,000Income Notes

Price

Trust

$8,2000,00012-year Treasury Zero Price

Price (Treasury Zero) + Price (Income Notes) = Par (Treasury Zero)

$10,000,000 ofIncome Notespriced at par

$18,2000,000 par of 12-year Treasury Zero, with

a price = $8,200,000

$18,200,000PPU

“Aaa”




Return Analysis

The vast majority of CDOs are privately placed. As a result, publicly available data on the historical performance of CDO bonds is scarce. This dearth of information makes it difficult to use traditional statistical techniques to analyze CDO income note returns and the volatility of those returns. The lack of historical information, however, does not mean that returns cannot be analyzed under a variety of stress scenarios.

An internal rate of return (IRR) of 15%–20% is often bandied about in the marketplace, but how robust and predictable is this return? What assumptions underlie such forecasted returns? The answers to these questions depend on the confluence of a number of factors, including:

➤ Magnitude and timing of defaults and sales at a discount to par.

➤ Magnitude and timing of recoveries.

➤ Interest rate movements.

➤ Calls and tenders.

Defaults Since most CDOs are repackagings of high yield corporate bonds and loans, their performance is correlated with the default trends in the high yield market. According to Peters and Altman,48 that default trend has been rising during the last few years, with aggregate defaults reaching 4.15% in 1999, up from 1.6% in 1998 (see Figure 82). Peters and Altman cite a number of factors contributing to the rising default rate, including heavy high yield bond issuance from 1997 to 1999, a trend toward defaults occurring a shorter period of time after issuance, deteriorating credit quality of new issues, and significant defaults in certain industries. Over the life of the study, however, defaults have been approximately 3% per year.

It is also important to note that although the performance of the CDO sector is correlated with high yield corporate sector in general, they are far from perfectly correlated. A CDO, after all, does not own all the credits in the high yield universe. It owns a carefully selected, diverse portion. Top-tier portfolio managers have proven that they can consistently experience lower defaults than the marketplace as a whole. As will be discussed in the “Asset Manager” section of this report, asset selection and the manager’s long-term track record are key.

48 Gregory J. Peters and Edward I. Altman, Defaults and Returns on High Yield Corporate Bonds, Analysis through 1999 and Default Outlook for 2000–2002, January 31, 2000.

The lack of publicly available CDO

performance data makes its difficult to use

traditional statistical techniques to analyze

CDO income note returns.

In the high yield market, defaults have averaged around 3%

since 1971.

Top-tier portfolio managers have proven

that they can consistently experience lower defaults than the

marketplace as a whole.



Figure 82. Historical Default Rate, High Yield Bond Market, 1971–1999

0

2

4

6

8

10

12

1972

1973

1974

1975

1976

1977

1978

1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

Defa

ult R

ate

(%

)


While the average annual default rate for the entire high yield universe has been about 3% over the last thirty years, the relatively small number of obligors in the typical CDO (i.e., 70–120) will result in default behavior that is considerably more volatile than any historical study. Consequently, we question whether the Street CDO equity pricing convention of applying a level 2% annual default rate is valid. It is impossible to predict which default pattern will prove to be the “right” pattern but it certainly will not be a smooth, steady 2% every year for the life of the income note. The prudent investor will take a view on future default behavior and test an income note under varying default assumptions before investing.

Unless otherwise stated in a particular figure, the base assumptions listed in Figure 83 will apply in all figures. Figures 84–85 illustrate the impact of various default scenarios on equity returns.

Figure 83. Cash Flow Modeling Assumptions Base line default rate 2%Recovery rate 50%Liability weighted-average cost of funds 10yr UST + 2.25%Asset yielda 10yr UST + 5.20%Fees 0.55%Reinvestment Rate 11%Interest rate hedge Notional amount equal to 90% of the initial asset baseaNet of management fees. Source: Salomon Smith Barney.

When assessing the characteristics of a

particular income note, investors should not

rely solely on the Street CDO equity pricing

convention of applying a smooth 2% annual

default rate.



Figure 84 is representative of the current equity pricing paradigm of applying a “smooth” default number over the life of the cash flow scenario. If a CDO’s annual defaults match the historical average of 3% described in the Peters and Altman study, the equity IRR would equal 14.8%, the PPU (Principal Protected Unit) IRR would equal 10.1% and an unleveraged investment in the pool would return 9.3%. As Figure 84 illustrates, the annual default rate must exceed 4% over the life of the CDO in order to make the unleveraged investment in the pool of bonds a better value relative to the leveraged equity investment. Interestingly, if annual defaults stay at a 3% rate, the PPU investment returns a paltry 0.8% above the collateral yield (10.1% versus 9.3%).

Figure 84. CDO Income Note Returns — Sensitivity to Annual Default Rates

7.9%

4.7%4.6%

8.9% 8.4%9.3%

10.7% 10.2% 9.8%

13.5%

8.2% 6.5%

10.1%

15.0%

11.8%9.8%

23.4%20.8%

17.9%

14.8%

-2.1%

-5%

0%

5%

10%

15%

20%

25%

0% 1% 2% 3% 4% 5% 6%

A nnua l Default Rate ( Cumula tiv e Default R ate )

IRR

C o lla te ra l PPU Equity

( 0% ) ( 7.5% ) ( 14.8 ( 21.4% ( 27.2% ) ( 33.2% ) ( 36.9% )


Figure 85. CDO Income Note Returns — Sensitivity to Default Rate Spikes

-3.8%

0.9%4.1%

7.4%

11.1%

14.6%

3.5%

7.3%

10.5%

13.4%

-0.5%

16.3%17.9%15.5%

13.4%11.4%

9.0%

6.1%

16.5%

-5%

0%

5%

10%

15%

20%

2% 4% 6% 8% 10% 12% 14%

S p ike

IRR

Y e a r 1 & 2 S p ike Y e a r 3 & 4 S p ikeY e a r 5 & 6 S p ike


If a CDO’s annual losses match the 3% historic

average, the equity IRR can equal a respectable 14.8%. An experienced

CDO manager can do even better.

The annual loss rate

must exceed 4% (27.23% cumulative) to make an

unleveraged investment in the pool of bonds a better value relative to

CDO equity.



Figure 85 illustrates the impact of default rate “spikes” on the equity IRR. Since small pools of corporate debt obligations have no predictable loss curve, default rate “spike” scenarios are key to understanding potential returns. If default rates remain in the 4% range (i.e., 33% above the historical average) for the next two years and then revert to 2% per annum, Figure 85 shows an IRR of about 15% (14.6%). If, over the next two years, default rates rise 50% above today’s average default rate (4.0% * 1.5 = 6.0%) and then revert to 2% per annum, the equity IRR in Figure 85 will equal approximately 11.1%.

As Figure 85 illustrates, loss avoidance in the early years is key.

Recoveries As default rates rose in 1999, recovery rates dropped.49 The average recovery after default was 28% in 1999, the lowest since 1990 and lower than both the 1998 recovery rate (36%) and the average recovery rate from 1978 to 1999 (42%); see Figure 86. This was the case even though less than 40% of the defaulted obligations were subordinated. One reason for the drop in recovery rates in 1999 may be the current glut of distressed and defaulted corporate paper.

Another contributing factor may be how some CDOs treat defaulted bonds. The way a CDO structure treats distressed and defaulted assets can have an impact on its overcollateralization (OC) test as well as on the ultimate recovery that the manager receives on the asset. Both of these can affect returns to the income note holders. For example, for the purpose of the OC test, defaulted assets typically are carried at the lesser of 30% or the market value of the asset. In addition, many structures require the manager to sell a defaulted assets at a maximum of one year after default. Distressed corporate market participants are all too aware of these artificial structural constraints and, as a result, we have seen bids for defaulted paper at or slightly above 30 cents on the dollar. Since CDOs have become such huge buyers of high yield paper and now hold, by our estimates, over 15% of the high yield market, their rules regarding the disposition of defaulted paper may have begun to affect the distressed bond market.

49 Ibid.

Since small pools of corporate debt

obligations have no predictable loss

curve, default rate “spike” scenarios are key to understanding

potential returns.

From 1978 to 1999, for all corporate debt

obligations, the average recovery rate immediately following

default was 42%.

CDO structural rules governing the

disposition of defaulted assets are putting

downward pressure on recovery rates.



Figure 86. Annual Recovery Rates, 1971–1999

0

10

20

30

40

50

60

70

80

1978

1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

Recove

ries

(%)


Lower sale prices drive recoveries lower and can ultimately diminish returns to the income note holder. For this reason, when examining an income note opportunity, an investor should put historical recovery studies in the context of the last two years. The recovery paradigm may be shifting. Figures 87–88 illustrate the impact of various recovery scenarios on equity returns.

Absolute recovery levels aside, all recovery studies show a tiering in recovery rates based on the defaulted instrument’s level of seniority and security.50 Historically, senior secured bank loans have shown the best recovery potential, followed in descending order by senior unsecured bank loans, senior secured bonds, senior unsecured bonds, and subordinated bonds. An investor should determine the potential asset mix of a CDO and the assets that are most likely to default before taking a view on the average recovery rate that the CDO may experience. Depending on the mix, the average recovery rate of 42% cited above may be too conservative or too generous. Either way, given the relatively small number of assets in a CDO, the actual recovery experience may differ from the averages cited in industry studies.

50 Karen Van de Castle and David Keisman, “Recovering Your Money: Insights Into Losses from Defaults,” Standard & Poor’s CreditWeek, June 15, 1999; and David T. Hamilton, Debt Recoveries for Corporate Bankruptcies, Moody’s Investors Service, Global Credit Research, June 1999.

The recovery paradigm may be shifting.



Figure 87. CDO Income Note Returns — Sensitivity to Recovery Rates and Annual Default Rates

15.6%

9.0%

1.7%

-11.4%

19.8%

20.8%23.4%

4.6%

9.8%

14.8%17.9%

21.9%

14.6%16.4%18.4%20.2%

-15%

-10%

-5%

0%

5%

10%

15%

20%

25%

0% 1% 2% 3% 4% 5%

Annual Default Rate

IRR

30% Recoveries 50% Recoveries 70% Recoveries


Figure 88. CDO Income Note Returns — Sensitivity to Recovery Rates and Default Rate Spikes

15.6%

-0.7%

5.0%

10.1%

17.9%

14.6%

11.1%

7.4%

15.0%

17.2%18.8%

20.2%

-5%

0%

5%

10%

15%

20%

25%

2% 4% 6% 8%

S pike — Y e a rs 1 & 2

IRR

30% R e c o v e rie s 50% R e c o v e rie s 70% R e c o v e rie s


Interest-Rate Risk Many CDOs are floating rate obligations backed by pools of fixed-rate bonds and, in order to hedge potential interest rate mismatch, most CDOs purchase a combination of interest rate swaps and caps. Hedging is one of the least standardized features of a cash flow CDO and, as a result, it must be analyzed on a deal-by-deal basis.

Most CDOs are not perfectly hedged because such a hedge would be prohibitively expensive. Incremental hedging costs are funded by the issuance of additional income notes, thereby diluting equity returns. Historically, as a result, most equity investors have been willing to accept some interest rate risk in exchange for returns that have not been dampened by excessive hedging costs.

Most CDOs are not perfectly hedged.



In a typical CDO, the trust pays a fixed rate of interest to the counterparty and the counterparty pays LIBOR to the trust. The swap notional amount amortizes pursuant to a schedule set at closing. Figure 89 assumes that the notional amount of a hedge is equal to 90% of the CDO capital structure and that the CDO liabilities are floating rate notes indexed to LIBOR.

Figure 89. CDO Income Note Returns — Sensitivity to LIBOR Movements

4.7%

22.7%

17.1%

13.9%

9.7%

0.3%

20.1%

-2.1%

4.6%

20.8%

9.8%

14.8%

17.9%

23.4%

-5.1%

4.2%

9.9%

15.6%

18.7%

21.6%24.1%

-6%

-2%

2%

6%

10%

14%

18%

22%

26%

0% 1% 2% 3% 4% 5% 6%Annual Default Rate

IRR

LIBOR +200bp Forward LIBOR Curve LIBOR -200bp


Call and Tender Premiums When a high yield bond is called or tendered, the premium paid to bondholders often equals half the annual coupon. Since many CDOs treat this premium as part of asset yield collections, increased call and tender activity can mean increased returns to the CDO income notes. We estimate that call and tender activity in the high yield market, while volatile, has averaged about 8.5% in the last seven years (Figure 90). Figure 91 demonstrates the effect of assuming a 2.5% and 5.0% annual call/tender rate. The figure assumes the calls/tenders are at a price of 105. At a 3% annual loss rate, a 2.5% call/tender rate can boost the equity IRR by more than 100bp.

Figure 90. Historical Calls and Redemptions (Dollars in Millions) 1993 1994 1995 1996 1997 1998 1999 AverageCalls/Redemptions 28,828 17,386 10,664 7,281 16,504 19,195 9,514Tender Offers 1,982 2,383 3,660 10,507 20,290 24,716 18,178Avge Size of HYM 213,537 237,297 261,453 301,453 363,917 453,047 534,361

Calls/Redemptions/Tenders as a % of HYM 14.43% 8.33% 5.48% 5.90% 10.11% 9.69% 5.18% 8.45%


LIBOR shifts can impact equity returns depending

upon how the liabilities amortize in relation to

the notional amount of the swap.

Although call and tender premiums can

boost returns by more than 100bp in some

scenarios . . .



Figure 91. Sensitivity to Call and Tender Premiums

11.35%

4.63%

9.77%

14.76%

17.92%20.85%

23.41%

6.49%

16.10%

18.99%

21.81%24.28%

7.32%

12.68%

17.32%

19.99%22.71%

25.12%

0%

5%

10%

15%

20%

25%

30%

0 1 2 3 4 5Annual Defaults

IRR

No Premium 2.5% Annual Premium 5% Annual Premium


Although it may be tempting assume call premiums when analyzing CDO equity, we recommend against it for a few reasons. First, call and tender activity is driven by a number of factors including changes in interest rates and company specific events. Over the 12-year legal life of the CDO, it is extremely difficult to predict the movement of interest rates and the effect of those interest rate changes on the high yield market as a whole, much less the impact on the 80–150 names that the CDO holds at any point in time. Also, many CDOs today contain a percentage of loans. Typically, loans may be prepaid without penalty at par at any time. Most importantly, while premiums offer some potential IRR upside, credit losses present a much larger threat on the down side. It would take a large number of calls at 105 to outweigh the impact of a few assets defaulting and being liquidated at 40, 20, or 10.

In addition to the numerous quantitative considerations explained above, there are a number of key qualitative factors that will determine relative value among income note investment opportunities. These factors fall into three main categories: (1) collateral manager; (2) asset characteristics; and (3) structural features. The remainder of this report will explore these qualitative factors in depth.

. . . potential income note investors should not rely

heavily on them when deciding whether to

make an investment.



Collateral Manager

Collateral Manager Review An arbitrage CDO is a hybrid-structured finance/corporate instrument whose performance is linked not only to the credit quality of the collateral and the nature of the structure but also to the portfolio manager’s trading decisions. The collateral manager’s initial asset selection and trading decisions throughout the reinvestment period are crucial.

The key attributes of a manager that investors should examine in depth are:

➤ Track record.

➤ Experience managing within the CDO framework.

➤ Level of institutional support.

➤ Investment and trading philosophy.

➤ Expertise in each asset class that the manager is permitted to invest in.

➤ Importance of CDO product to overall organization.

➤ Manager’s access to assets.

An asset manager review is the best way for an income note investor to get a firm grasp of a manager’s strengths, weaknesses, and historical performance. Some key discussion points and portfolio performance information requirements are listed in Figures 92 and 93.

Figure 92. Collateral Manager Review Check List Company Overview Financial strength of the company Experience in corporate lending and managing portfolios of high yield bonds and bank loans How does managing a CDO fulfill the company’s strategic objectives? Importance of CDO to overall organization Prior history managing CDOs Is entire CDO managed by a couple of key decision makers (“key person” risk)? Number of high yield funds under management Performance results relative to peer group and index benchmarks Compensation arrangements for the collateral managers Will the company purchase part of the CDO income note?

Research Research methodology Industries covered Number of analysts and credits per analyst Depth of analyst contacts with industry participants Ability to expand research to cover additional industries required in a diversified CDO Sample research reports

Underwriting and Investment Strategy Credit and approval policy Investment style Facility with and understanding of bond indentures and loan covenants Decision making process for buy and sell decisions Pricing sources and policies regarding securities valuation

Credit Monitoring Procedures to service the CDO and to ensure compliance with the CDO transaction documents Does the manager have in-house cash flow modeling capabilities or does it rely solely on the trustee and/or underwriter? Frequency of credit reviews Technological tools used to monitor the portfolio Procedures for managing credit-risk and defaulted assets

Sources: Standard & Poor’s, Fitch IBCA, and Salomon Smith Barney.

The collateral manager’s initial asset selection and trading decisions

throughout the reinvestment period are

crucial drivers of CDO income note returns.



Figure 93. Portfolio Performance Defaults and Credit-Risk Sales Default history Credit risk sales below 80 Asset specific rationale for each credit risk sale Where has each credit-risk asset traded after sale? Did it ultimately default? Length of time between an asset purchase and its sale as a credit-risk asset

Recoveries Recovery history Method of disposition: sale after default or buy-and-hold Recovery timing

Trading Annual turnover rate for the portfolio Frequency of credit-improved sales Credit-improved sales: average premium to purchase price

Returns Compare annual returns to peer group and index benchmarks Volatility of annual returns


Asset Selection A CDO manager can outperform the market depending upon which names and industries it chooses initially and the trading decisions it makes during the reinvestment period.

Investors should be aware, however, that the time period during which the collateral manager purchases its collateral (“cohort” or “time stamp”) can have as much effect on the performance of a CDO as the skill of the manager. For example, as a general matter, CDOs which ramped-up during the spring of 1998 have not performed as well as other CDOs. During early 1998 asset spreads were tight and managers had to venture down the credit curve and invest in marginal credits in order to generate sufficient returns to CDO equity investors. One way CDO equity investors can mitigate the potential risk associated with cohort to identify a list of approved, “blue-chip”, CDO managers and invest serially in CDOs issued by those managers.

Time stamp aside, we agree with the thesis that the high yield market is not as efficient as other markets and, as a result, there are opportunities for CDO managers who are well versed in fundamental credit analysis and have access to timely information to outperform their competitors.51 The high yield market does not price every asset accurately. An experienced manager knows which assets are cheap relative to default probability and which are priced properly. Those CDO managers with strong research teams, good industry contacts, robust deal flow, and sophisticated systems have a good chance of outperforming the market.

Prudent asset selection is crucial because the asset pool supporting a typical arbitrage CDO is granular. Although the trend is towards larger pools and smaller obligor concentrations, the typical high yield arbitrage CDO may have as few as 70–120 names. With obligor concentrations ranging from 1% to 3%, a handful of poor investment choices may substantially reduce returns to income note holders.

51 Fitch IBCA, Management of CBOs/CLOs, Robert J. Grossman, December 8, 1997

The time period or “cohort” during which the manager

purchases the collateral can have as much effect on the performance of the CDO as

the skill of the manager.

Prudent asset selection is key because the

asset pool supporting a typical arbitrage CDO

may contain as few as 70–120 names.



CDO Investment Guidelines ➤ A good total return high yield manager does not necessarily equate to a good CDO

manager. Two major differences exist. Total return arbitrage CDOs are typically leveraged 8–12 times, and this leverage greatly magnifies returns and losses to the CDO income notes. Also, managing within a CDO’s arcane and cumbersome investment guidelines (which have been crafted to garner investment-grade ratings on the senior notes from the rating agencies) can be challenging.

In a typical CDO, a manager must satisfy more than twenty investment guidelines before making a trade. No trade is easy. Figure 94 illustrates the guidelines which must be satisfied before a manager can make a purchase.

Figure 94. Typical CDO Investment Guidelines Minimum average asset debt rating Minimum percentage of assets rated B3 or better Minimum percentage of assets in U.S. Maximum percentage of assets outside U.S., Canada and U.K. Maximum percentage of synthetic securities Minimum diversity score Maximum single issuer exposure Maximum percentage in any S&P industry group Maximum percentage in any Moody’s industry group Maximum percentage of zero coupon bonds Maximum percentage of loan participations Maximum percentage of floating rate securities Weighted average life test Class A, B, and C minimum O/C tests Class A, B, and C minimum I/C tests Minimum weighted average recovery test Maximum percentage of securities maturing after a certain date Minimum average asset margin test Minimum average asset coupon test Maximum annual discretionary trading bucket


Given the complexity of these investment guidelines, a manager who currently manages one or more CDOs will have a distinct advantage over a first-time CDO manager, all other factors equal. Although a new CDO manager may have a conceptual understanding of each of these guidelines in isolation, until a manager operates within them and understands the interrelationship among all guidelines, they are difficult to master.

If a CDO collateral manager has mastered the investment guidelines within a CDO, in what ways may this benefit the income note holders? One way involves industry diversification. All rating agencies encourage industry diversification for the benefit of the rated note holders. A manager must seek the optimal amount of industry diversification: diversification that maximizes the rating agency “credit” to rated note holders, minimizes forays into unknown industries, and generates a fair risk-adjusted return to income note holders.

Another way that an experienced CDO manager can benefit income note holders is by taking a balanced view of a CDO’s asset eligibility parameters. Many modern CDOs allow a manager considerable flexibility to invest in various nontraditional assets, such as emerging market debt and structured finance obligations. These asset types may generate significantly more yield than comparably rated high yield bonds.

A good total return high yield manager does not necessarily equate to a

good CDO manager.

CDO investment guidelines are myriad

and cumbersome.

A manager who is currently managing one

or more CDOs will have a distinct advantage over a first-time CDO manager.

An experienced CDO manager will take

a balanced view toward industry and

obligor diversification.



They present an enticing way for a manager to “juice up” returns to income note holders. But, as with the perils of industry diversification, a manager who is too aggressive in searching for additional yield in nontraditional products may invest in asset types that it does not understand. Enhanced returns to the income note holder may not provide adequate compensation for the additional risk. A prudent CDO manager will resist this urge and instead stick to asset classes that it understands, even if that means forgoing some yield opportunities. In the long run, this should ensure a more stable, risk-adjusted return to the income notes.

CDO Manager Types CDO managers run the gamut from giant, highly rated banks and insurance companies to small, specialized, high yield portfolio managers. The size of the CDO manager does not, on its own, determine whether a particular income note is a good investment opportunity. For example, a CDO business that is a tiny part of a large insurance company or bank may not receive the same level of attention as a CDO business which is managed by a high yield boutique. In the latter case, the success of the CDO business is crucial to the success of the business as a whole. On the other hand, if an insurance company sponsored CDO falters or a key portfolio manger departs, the insurance company will have greater financial wherewithal to support the CDO business and hire a capable replacement.

An income note buyer should also understand how the portfolio manager makes its investment decisions. An institution that centralizes its investment decisions with one or two key people runs the risk that those key people might leave the institution for other opportunities. For this reason, income note investors should try to ascertain whether the sponsoring institution espouses an investment philosophy and whether this philosophy is shared by a broad cross section of the CDO management team. These team members should participate actively in all trading decisions.

Another key indication of support is whether an institution has issued multiple CDOs. If so, this indicates an institutional commitment to the CDO business. That commitment is further strengthened if the institution is an income note investor in each of its CDOs.

Investment and Trading Philosophy A CDO manager’s investment philosophy and trading style will have a significant impact on returns to the income notes. A key indication of this style is how the manager strikes a balance between the rated notes and the income notes. Although rated note holders and income note holders share many of the same concerns, their interests diverge in some important ways. Their viewpoints often differ as to the optimal investment and trading philosophy for a CDO manager.

Rated note holders occupy the majority of the capital structure of the CDO, and their primary concern is the preservation of principal and a coupon entitlement that is attractive relative to other similarly rated fixed-income instruments. These note holders are concerned with initial asset selection before closing and during the ramp-up period. Once a transaction is ramped up, triple-A note holders are averse to a CDO collateral manager that actively trades the portfolio because they rely upon

The size of the CDO manager does not, on

its own, determine whether a particular

income note is a good investment opportunity.

A manager that takes a collaborative approach

to asset selection is less susceptible to key

person risk.

A manager that has issued multiple CDOs and has

purchased equity in each has indicated an institutional

commitment to the business.

How will a manager strike a balance

between the interests of the rated note

holders and income note holders?

Rated note holders rely on asset cash flow, not

trading gains, to service their debt.



asset cash flow, not trading gains, to service their debt. As long as assets do not default, they will produce the necessary cash flow to service the rated debt. The market value of the underlying asset pool may be of interest as a leading indicator of credit quality, but it is not of primary importance to rated note holders.

Income note holders do not think in terms of preservation of principal and a fixed coupon payment. They think in terms of cashflow and a return on their initial investment. As we have explained, this return is driven by, among other things, defaults, recoveries, interest rates, premiums, and trading gains and losses. Like rated note holders, income note holders focus on a manager’s initial asset selection, because prudent asset selection can minimize losses and benefit all note holders. Unlike rated note holders, however, income note holders are concerned with the market value of the assets in the CDO and the manager’s trading decisions, if any, regarding assets that are trading at premiums and discounts.

There are three categories of trades that a manager can make: credit-risk, credit-improved, and discretionary. Rated note holders and income note holders often have differing views as to the advisability of a particular trade. If a manager has a CDO or CDOs outstanding, a potential income note investor can analyze the manager’s past trades and infer whether the manager has worked to preserve principal for all note holders or has concentrated on enhancing returns to the income note holders.

➤ Credit-risk sales. A credit-risk sale is a sale of an asset that has declined in credit quality and that the manager reasonably believes will default with the passage of time. A credit-risk asset is sold at a discount to par and this sale, in isolation, results in the reduction of the asset base supporting the notes. This loss of principal will move the actual overcollateralization (OC) closer to the minimum OC trigger. If the minimum OC trigger is tripped, collections will be used to pay down the senior notes.

Rated note holders will view credit-risk sales favorably only if: (1) the asset ultimately defaults; and (2) the sale price is greater than ultimate recovery on the defaulted asset. If an asset is sold as credit-risk and does not default before the CDO is retired, the CDO has taken a loss (i.e., sale at discount to par) that it could have avoided if the asset had been held to maturity.

Income note holders may have differing views concerning credit-risk sales. Some may prefer managers to hold onto credit-risk assets because any discounted sale would push actual OC closer to the minimum OC trigger and increase the risk of de-levering. Other income note investors may prefer early aggressive sales of credit-risk assets at slight discounts rather than waiting for an asset to trade at a steep discount.

➤ Credit-improved sales. Credit-improved sales can benefit both rated and income note holders. The issue hinges upon how the premium is treated in the structure. The premium generated from a credit-improved sale may be treated as interest collections and used to enhance returns to the income note holder or it can be used to grow OC through the purchase of additional assets. Clearly, the latter method benefits all note holders. When a manager sells an asset that has improved in credit quality, the rated note holders lose the benefit of upward credit migration but if the sale proceeds (including premium) are reinvested in additional assets, the manger may be able to maintain or increase OC.

A manager’s trading patterns in existing

CDOs can give potential investors important clues as to how the

manager has balanced the interests of rated

note holders and income note holders.



➤ Discretionary sales. Depending on the structure, a CDO manager also has the discretion to trade 10%–20% of the portfolio annually. Not surprisingly, rated note holders have a bearish view of unfettered discretionary trading: they prefer managers with strong credit fundamentals to execute a long term investment strategy. Any problem credits can be traded under the credit-risk trading rules. In contrast, income note holders favor discretionary trading provisions, because these allow the manager to continually search for assets with the best risk-adjusted returns.

In the final analysis, trading is a two-edged sword. Trading can expose cash flow CDO note holders to market value risk, but it also can be used to improve the credit profile of the pool of assets and could ultimately be a very positive force in mitigating credit risk.



Asset Characteristics

Collateral Mix In addition to high yield bonds and bank loans, arbitrage CDOs increasingly include nontraditional investments (see Figure 95).

Figure 95. Nontraditional Assets Loan participations Emerging markets sovereign debt Emerging markets corporate debt Distressed debt Credit derivatives Structured finance obligations Convertible bonds Mezzanine loans with warrants Project finance loans and bonds


Most CDOs have certain limitations or “buckets” for these types of assets, but the limitations are different for each CDO. Such assets are often included in arbitrage CDOs because their generous yields enhance arbitrage opportunities for the collateral managers and, ultimately, the income notes. Although these nontraditional assets offer enticing yield pickup, the income note holders must be certain that the collateral manger has sufficient investment experience in the particular asset class. If not, enhanced short-term income note returns may be outweighed by significant long-term credit risk.

The inclusion of nontraditional assets also raises a credit question. The credit risk inherent in all cash flow CDOs is analyzed using various corporate default studies. Since these are studies of corporate instruments they are not directly applicable to assets like structured finance obligations and project finance loans. Some have argued, however that applying corporate default studies to structured finance instruments is overly conservative, given that there have been far fewer structured finance defaults than corporate defaults over the last ten years.

Some nontraditional investments can be categorized as “bivariate risk” assets. These include loan participations, emerging market corporate debt, and credit derivatives. With respect to each of these assets, the CDO is exposed to the nonperformance risk of more than one counterparty. For example, if a collateral manager invests in a credit derivative, the CDO will not receive payment if either the underlying referenced obligation or the credit derivative counterparty fails to perform. Most CDOs allow a manager to invest up to 20% of a CDO’s assets in bivariate risk assets, but many managers do not avail themselves of this opportunity. If the collateral manager plans to utilize the 20% bivariate risk bucket or has used it in past transactions, the income note holders should determine whether they are being compensated for this additional risk.

Nontraditional asset classes offer enticing

yield opportunities as long as the manager

has investment experience in the

particular asset class.



Finally, the inclusion of nontraditional asset types may have an impact on assumed recovery values. Over the last few years, several large recovery studies have been completed, but each revolves around defaulted US corporate bonds and loans. If a manager aggressively invests in sovereign debt or structured finance obligations, the applicability of these studies becomes questionable.

Time Stamp or Cohort The period of time during which a CDO is ramped up can have a significant impact on its long-term performance. Depending on market conditions, the collateral manager will purchase assets from both the primary and secondary market. Historically, a large percentage of the assets (10%–50%) emanates from the new-issue calendar, and during the average ramp-up period (3–6 months), that calendar contains a finite number of names. Consequently, arbitrage CDOs that are ramped up during the same period may share a large percentage of the same names. Accordingly, if an investor purchases multiple income notes from CDOs that have concurrent ramp-up periods, there is the risk that the performance of these income notes may be correlated. This risk will decline after the end of the ramp-up period as the manager starts trading the portfolio and the risk may not be as pronounced if one CDO manager is purchasing loans and the other is purchasing high yield bonds.

The prices of high yield bonds and bank loans during the ramp-up period can also have a big impact on the performance of a CDO. As we mentioned earlier in the report, during the fall of 1998, prices for high yield bonds dropped and spreads widened considerably for technical reasons, although underlying credit fundamentals were relatively stable. CDO managers that purchased collateral during that time frame were able to buy good credit quality collateral at discounted prices. Discounted prices allowed managers to purchase much more collateral than they had projected without going down the credit spectrum. Many of these deals, consequently, have asset buffers that are significantly above their minimum overcollateralization tests. These managers did not time the market: it was fortuitous that they came to the market during that period.

For these reasons, if an investor is going to build a portfolio of income note investments, we recommend that it purchase income notes that are issued during different time periods or cohorts. Investors can execute this strategy in two ways. They can review the new-issuance calendar for the next quarter or two and select income notes from various CDO issuers. This would give them maximum exposure to different credits and CDO manager investment styles. Alternatively, since the performance of CDO income notes is tied so closely to the skill of the manager, an investor may approve certain “blue-chip” managers and buy income notes from each of their deals over time. An investor who chooses the second strategy will likely be exposed to some of the same credits across all CDOs that a manager issues. A manager tends to buy additional exposure to names it likes.

Arbitrage CDOs that are ramped up during the same time period may

share a large number of the same names and,

consequently, their performance may

be correlated.

During the ramp-up period, the relative

richness or cheapness of the underlying high

yield market can have a long-term effect on the performance of a CDO.



Diversification The rating agency methodologies encourage obligor and industry diversity. The theory is simple: since CDO asset pools are lumpy to begin with, the more names in the pool, the less any one obligor default can hurt note holders. Similarly with industries, if one industry is experiencing higher than average defaults, note holders’ exposure to that industry is limited. Rated note holders favor broad diversification because they are interested in preservation of principal and the payment of a fixed coupon. Income note holders are less sanguine about zealous diversification because diversification, while limiting credit risk, also limits upside opportunities. The income note holder wants the manager to make a few right “picks” that can have a disproportionately beneficial impact on income note returns.

How does the manager strike a balance between the interests of the rated note holders and the interests of income note holders? At some point, too much diversity can work against all note holders. No note holder benefits if overly restrictive CDO investment guidelines force a manager to invest in obligors and industries that it does not fully understand. Credit risk increases and income note returns decline.



Structure

Unlike certain structured finance products (e.g., credit card ABS), CDO structures are far from commoditized. Every CDO underwriter uses a different base structure, and even CDOs underwritten by the same banker can contain significant structural variations that can affect the income note holder. Income note investors who study these features in each CDO before deciding to invest may be able to deduce how the manager intends to strike a balance between the interests of the rated note holders and the interests of the income note holders.

The structure of a CDO is an important consideration for the income note holder because the income notes are structurally subordinated to the other notes issued by the CDO. From a cash flow perspective, the income note holder is not entitled to cash flow until payment of: (1) all fees and expenses (capped and uncapped); (2) interest and principal to more senior notes; and (3) all hedging costs (including termination payments). If these obligations have been paid and the minimum interest coverage (IC) and overcollateralization (OC) tests are in compliance, the income notes are eligible for distributions.

Trigger Levels All arbitrage CDOs contain two types of coverage tests: an asset coverage test (minimum OC test) and a liquidity coverage test (minimum IC test). If these tests are violated, reinvestment of principal ceases and principal and interest collections are used to accelerate the redemption of the senior notes until these tests are brought back into compliance. These triggers function as structural mitigants to credit risk. Because violation of these coverage tests can result in the payment of all cash flow to the senior note holders (and consequently none to the income note holders), income note holders should have a firm understanding of how they function.

One of the key ways to gauge the robustness of a projected IRR is to compare the actual OC and IC in the transaction to the minimum IC and IC triggers set by the collateral manager and deal underwriter. If the difference between actual and minimum is small, the triggers have been structured “tightly” by the collateral manager and the deal underwriter in an effort to give the CBO issuer a higher degree of leverage (i.e., enhance the projected IRR to the income note). If the relationship between actual and minimum is larger, the triggers have been structured more “loosely.” Although it may allow an underwriter to present a higher IRR to potential income note investors, a tight trigger is easier to violate and thus makes the IRR potentially more volatile.

An income note investor should also explore the relationship between the actual levels of OC and IC and the trigger points in the context of the overall credit quality of the portfolio. A portfolio with an average credit quality of single-B should, all other factors equal, have a larger income note and less leverage (as a percentage of the deal) than a portfolio with an average credit quality of double-B. Also, the CDO supported by the single-B portfolio should have a larger buffer between the actual OC level and the minimum OC trigger, since single-B default rates are more volatile than double-B default rates.

Although all CDO structures share certain

structural elements, each CDO structure is unique.

CDO income notes are structurally

subordinated to the other notes in the CDO

capital structure.

The IC and OC tests are structural mitigants to

credit and liquidity risk.

Are the triggers “tight” or “loose”?

A tight trigger is easier to violate and thus

makes the IRR potentially more volatile.



Finally, in most CDO structures, each class has its own minimum OC and IC test and the tests associated with the most subordinated rated class should trigger first. Nevertheless, the income note investor should analyze cash flow runs to understand under a variety of stress scenarios which tests trigger the pay down of the deal.

Senior Costs, Swaps, and Caps Portfolio management fees and the coupon payable to the rated note holders are two costs that can affect the cash flow payable to the income notes. An income note holder should examine the manager’s fee in each CDO and compare it to fees payable in other arbitrage CDOs. A typical fee structure will pay the manager 0.25% prior to payment of interest on the rated notes and at least 0.25% after payment of fees and rated note interest and the satisfaction of the IC and OC tests.

More importantly, as we have described in the Return Analysis section of this report, in many arbitrage CDOs the assets are primarily fixed-rated bonds and the liabilities are issued as LIBOR floaters. These deals typically use a combination of swaps or caps to hedge interest rate risk. The swaps and caps usually have notional amounts that amortize on a predetermined basis. This presents the risk that the transaction may be underhedged or overhedged at any point in time (see Figure 89). If the deal is underhedged, for example, more of the asset cash flow will be used to meet rated note debt coverage and less will be available for the income notes. Moreover, these hedges can terminate, and if the SPV owes a termination payment to the counterparty the payment will be made senior to payment of any residual cash flow to the income notes. Since termination payments can be large, investors should analyze the swap documents for each deal and understand which events can cause the termination of the swap.

Manager Fees and Equity Ownership There are a few ways that a structure can more closely align a portfolio manager’s economic interests with those of the income note holders. One way is through the payment of the portfolio manager’s fee. In some older transactions, the manager’s fee is paid before payment of rated note holder interest. This senior position is beneficial if the CDO needs to attract a replacement manager but it does not align the interests of the manager and the income notes. Even if the CDO is performing very poorly, the manager still gets paid its full fee. For this reason, most recent deals pay part of the fee at the top of the waterfall (base management fee) and part of the fee after payments of other fees, rated note holder interest and the satisfaction of the IC and OC tests (performance management fee). By subordinating a portion of the manager’s fee, these structures encourage the manager to generate enough cash flow to service the rated debt in a fashion that preserves the asset base and does not violate the IC and OC tests. Some structures pay the manager an additional fee if the actual IRR paid to the income note holder hits a certain target.

Another way a manager can align its economic interests with those of the income note holders is by purchasing a portion of the income note. This is the case in most CDOs. The theory: since the manager owns part of the income notes, it will manage the portfolio so as to produce reasonable returns to the income notes while protecting

At any point in time, a CDO will be

either overhedged or underhedged.

Swap termination payments are senior to

income note distributions.

A split-fee structure and equity ownership

more closely aligns the economic interest of the manager with the

economic interests of note holders.



them from unreasonable credit risk. Although many deals do not explicitly prohibit the manager from selling its portion of the income notes, as a practical matter the market for income notes is limited. In all likelihood, if a manager purchases income notes, it will retain them.

Credit-Improved Sales — Treatment of Premium

CDO investment rules allow a portfolio manager to sell an asset that has improved in credit quality and is now trading at a premium (credit-improved sale). What is a credit-improved sale and how are sale proceeds distributed? Definitions vary. Some structures define a credit-improved sale as a sale of an asset that has improved in credit quality and can be sold at a premium to purchase price. Other CDO structures describe a credit-improved sale as a sale of an asset that has improved in credit quality and can be sold at a premium to par.

CDO structures treat gains differently. Some treat premiums as principal proceeds that will be reinvested in new collateral. Rated note holders favor this treatment because premium sale proceeds are used to buy more collateral and enhance overcollateralization in the structure. Some income note holders may favor this treatment for the same reason. Other structures treat premium sale proceeds as interest proceeds that can be distributed to the income note holders if fees and rated coupon have been paid and the IC and OC tests have been satisfied. Rated note holders do not favor this version, because it allows a manager to skim all the credit upside off the pool of assets and stream it to the income note holder in the form of an enhanced IRR. Still other structures give the manager the option of designating premium proceeds as either interest or principal. Finally, another variation weighs the cumulative losses against the cumulative gains that a manager has incurred over the life of the CDO. If cumulative losses exceed cumulative gains, the proceeds of any credit-improved sale are deemed principal proceeds.

Some structures reinvest premium proceeds in additional assets and

some characterize premium proceeds as additional yield

eligible for distribution to equity holders.



Conclusion

During the last few years, demand for CDO equity has broadened substantially from large institutional investors to other investors such as small pension funds and high-net-worth individuals. A CDO equity investment program that purchases income notes from a select group of experienced CDO managers across various periods of time can be a effective way for an investor to diversify its portfolio and improve its risk-adjusted returns. We expect that continued growth in the CDO market will drive increased demand for CDO equity investments in the United States and in overseas markets, including Europe, the Middle East, and Asia.

JULY 2000

Glen McDermott Terry Benzschawel Adrian Lui

CDO Combination Securities Putting the Pieces Together



Su

mm

ary

Traditional Collateralized Debt Obligation (CDO) investment

structures may not be flexible enough to meet the objectives of all

investors. CDO combination securities have arisen to address this

need. Combination securities can be tailored for each investor

based on the desired credit rating, minimum coupon, yield target,

and capital guidelines.

Combination securities can also be used to provide downside protection to CDO equity tranche exposure with some sacrifice of upside potential. We quantify the trade-off between downside protection and this upside limitation using conventional constant default rate (CDR) methodology and Salomon Smith Barney’s Monte Carlo Simulation Method as introduced in our report on investment-grade CDOs.52

For various combinations of rating, rated coupon, and percentage of equity tranche, we compute internal rates of return (IRRs) using the CDR method, and calculate both IRRs and Sharpe ratios using our simulations. Although both methods predict similar returns for the various combinations, only the Monte Carlo method allows us to calculate the relative risk-reward characteristics of each instrument.

Results confirm our earlier analysis (see Footnote 52) that securities (equity and otherwise) backed by pools of investment-grade bonds produce high expected returns and high Sharpe ratios under historical default scenarios. For example, our prototypical CDO equity has an expected IRR of 18% with a Sharpe ratio of 1.7, whereas an Aa2 rated combination security with 8% equity and a minimum rated coupon of 4% yields 8% with a Sharpe ratio of 5.0. Simulations based on 2x historical default rates decrease the yield on the CDO equity to 15% and its Sharpe ratio to 0.6, whereas the combination security’s yield decreases to 7.4% with a Sharpe ratio of 2.3.

In general, we find a direct relationship between the expected IRR and the proportion of CDO equity in each security, but an inverse relationship between CDO equity and their Sharpe ratio. In addition, not all similarly rated combination securities produce equal expected returns and, in particular, may differ greatly in Sharpe ratios.

For any combination security an investor must decide how much yield to sacrifice for stability of return. Results from our sample of combination securities allow investors, based on their yield requirement and volatility tolerance, to determine the appropriate mix of senior, mezzanine and equity tranches supporting their investment.

52 McDermott, G., Benzschawel, T., and Khan, S., Investment-Grade CDOs, Salomon Smith Barney, August 2001.



Introduction

Combination Securities Defined The CDO market has provided investors with the ability to capture the benefits of a diversified portfolio of corporate bonds. The CDO, via its senior, mezzanine and equity tranche structure, allows the investor to participate in this diversified pool of assets at different risk-reward levels. For example, while the typical Aaa-rated CDO tranche in the market today may pay an investor LIBOR + 45bp, an equity interest in the same CDO has an expected yield of 15%–20%. However, the three-tiered investment structure may not be flexible enough to address the objectives of many investors.

For this reason, the market has developed a class of structures called combination securities, whose return is derived from the cash flows of two or more underlying instruments, at least one of which is a CDO security. The number of possible combinations is infinite but the creation of any combination security starts with the simple question: What are a buyer’s key investment considerations?

Important investment guidelines include:

➤ Does the investor require a rating on the security?

➤ Does the investor require a rated coupon or yield on the security?

➤ What is the investor’s “economic” yield target?

➤ What are the investor’s economic and regulatory capital guidelines?

➤ Under what legal jurisdiction does the investor operate?

Principal Protected Units A combination security may be crafted to satisfy almost any investor requirement. Figure 96 illustrates the simplest form of combination security, the principal protected unit (PPU).

Figure 96. Principal Protected Unit Diagram

$22,000,000Price of CDO Equity

Trust

$45,000,000PPU

"Aaa"

$23,000,000Price of 12-year

Treasury Zero with$45MM Notional


A principal protected unit is created by bundling a piece of CDO equity with a zero coupon Treasury STRIP (or other security free of credit risk). These units are designed to protect investors from the loss of their initial investment while still providing the potential for attractive returns. In Figure 96, the buyer invests a total of $45 million for

Traditional CDO investment structure may not be flexible enough to address the objectives of

many investors.

The return on CDO combination securities is

derived from the cash flows of two or more

underlying instruments, at least one of which is a

CDO security.

The simplest form of combination security is the principal protected

unit (PPU).



the PPU, $23 million of which is used to purchase the 12-year STRIP with a face value of $45 million, and the remainder of which is used to purchase CDO equity. Even if the equity fails to return any cash flow, the STRIP will accrete to a face of $45 million at maturity, and the investor will recoup the initial capital outlay. Moody’s can rate this PPU triple-A for return of principal, which may allow some investors to gain preferable capital treatment under internal and/or external capital guidelines.53

The cost of this downside protection, however, is to sacrifice some of the upside potential. Since both CDO equity and a Treasury STRIP support the PPU, the STRIP will dilute expected returns and limit the return of the PPU when compared to an investment solely in the equity. Figure 97 compares expected returns on a CDO equity investment to a PPU investment in an investment-grade average CDO.54

Figure 97. Expected IRR from CDO Equity Versus Principal Protected Unit

-25%

-20%

-15%-10%

-5%

0%

5%

10%15%

20%

25%

0.00% 0.25% 0.50% 0.75% 1.00% 1.25% 1.50% 1.75% 2.25%Annual Default Rate

IRR

Equity

Principal-Protected Unit


The average credit quality of the typical portfolio that supports an investment-grade CDO is Baa3, and the average one-year default rate since 1983 for Baa3 credits is approximately 0.5%. Assuming the manager of the investment grade CDO in Figure 97 matches this average default rate, the return on the equity would be 17.1% as compared to a 9.3% return on the PPU — under this scenario, the investor has forgone about 800 basis points of return. On the other hand, if the portfolio’s one-year default rate is three times the historical average since 1983 (i.e., 1.5%) for the life of the transaction, the PPU (5.0%) will outperform the equity (3.1%). This analysis does not imply that one of these investments is superior to the other, but one may be more appropriate for an investor depending on their risk, yield and rating requirements.

The Evolution of Combination Securities In recent years, combination securities have evolved from the simple PPU to more sophisticated combinations. Figure 98 illustrates the latter. In this figure, the return characteristics of a “plain vanilla” Baa2 CDO tranche are compared with the returns

53 For instance, European insurance and pension funds face legal restrictions on the amount of foreign currency in their investment portfolios. Since 80% of a European insurance company’s investment assets must be in euros, their dollar- and yen-denominated investments are limited. Providing a PPU containing a euro STRIP increases the investor base with the ability to participate in CDO equity. Therefore, the investor takes currency risk only on the return portion of their investment and not on the principal portion.

54 These and other scenarios generated in this paper are based on many assumptions, including a 50% immediate recovery rate for defaults.

The cost of the Treasury STRIP dilutes

PPU returns.

Combination securities have become increasingly

more sophisticated.



on a Baa2-rated combination security issued from the same CDO.55 A Baa2-rated tranche (30%) and equity (70%) from the underlying CDO back the combination security and the Moody’s rating reflects return of principal only.

Figure 98. Expected IRR of a Baa2 Combination Security Versus a Baa2 Plain Vanilla Note

0%

3%

6%

9%

12%

15%

18%

0.00% 0.25% 0.50% 0.75% 1.00% 1.25% 1.50% 1.75% 2.25%

Annual Default Rate

IRR

Baa2 Plain Vanilla Note

Baa2 Combination Security


These two securities have the same rating (and hence, the same expected loss) but substantial difference in the return volatilities. The return profile for the plain vanilla CDO tranche remains unchanged under all but the most stressful default scenarios — the 8% return deteriorates only slightly (7.3%) when defaults hit 2.25% per annum.56 By contrast, the Baa2 combination security can range from 16.4% with zero defaults to -0.4% with 2.25% annual defaults. Under the traditional method of analyzing combination securities, an investor will assume an annual default rate based on the manager and the structure. In this example, if the manager experiences defaults at a rate of 0.5%, the combination security return (13.8%) exceeds the return on the plain vanilla note (8.0%) by 5.8%. If defaults hit 1.5% per year, however, the combination security (5.7%) will underperform the plain vanilla note (8.0%).

The results in Figure 98 appear to indicate that, at least for historical default rates, the combination security is a better investment than the plain vanilla Baa2 security. However, in our opinion, the ratings-based, constant default rate (CDR) investment framework has limitations. First, as Figure 98 illustrates, two securities can have the same rating but radically different return profiles. Second, the CDR method assumes that default rates are uniformly distributed in the investment-grade and high yield rating categories. They are not. Third, combination security returns are critically path dependent (i.e., defaults occurring earlier in the life of a transaction will impair an investment more than defaults occurring at the tail end). Finally, while the traditional methodology produces average returns for a given annual default rate, it gives no guidance to an investor regarding the dispersion (i.e., the variability) of returns around that mean.

For these reasons, in the Methodology section of this paper, we propose an improvement upon the market convention. First, though, we summarize Moody’s expected loss methodology and how it has been extended to combination securities.

55 By “plain vanilla” we mean a standard, non-combination CDO tranche. The plain vanilla Baa2 rated CDO tranche was analyzed assuming a coupon of 8%. For the purposes of this example, we used an actual, recent vintage investment-grade-average CDO.

56 A 2.25% per annum default rate is extremely conservative for investment grade CDOs. The average credit quality of these portfolios is Baa3 and, as we have mentioned previously, the average one-year default rate since 1983 for Baa3 credits is around 0.50%. A 2.25% per annum default rate would be 4.5x the average for the life of the transaction.

Two CDO securities can have the same rating

but substantially different return profiles.

The ratings-based, constant default rate analytical framework is limited.



Ratings of Combination Securities

Moody’s Expected Loss Methodology The Moody’s rating of all CDO securities, plain vanilla and combination, is based upon its expected loss methodology (ELM). In this section, we will briefly describe ELM and its extension to combination securities. Under ELM, the lower the expected loss for a tranche, the higher its rating.

The first step in determining the expected loss for a tranche involves the computation of cash flows, whereby a series of increasingly severe default scenarios are applied to the collateral portfolio and the loss (Ls) attributed to each CDO tranche under each scenario (s) is weighted by its corresponding probability of occurrence (pr(Ls)). The product from each of the default scenarios is then aggregated to quantify the expected loss (E) for the tranche.57 This estimate of the expected loss is compared to Moody’s “idealized” loss table to arrive at a Moody’s rating for the CDO security.58

)]([0

s

D

ss LprLE ∑

=

×=

The loss on a CDO tranche in any given default scenario is calculated by comparing (a) the present value of its projected cash flows using a discount rate equal to its rated coupon and (b) its initial principal amount. If (a) equals (b), the CDO tranche is able to pay off its stated coupon and principal and suffers no losses (loss of zero). If (a) is smaller than (b), the net present value (NPV) is negative and the CDO tranche suffers a loss. The loss for each scenario is calculated as follows:

bab

L ss

),0max( −=

Extending ELM to Combination Securities As we have mentioned previously, the rating methodology applied to combination securities is the same as that applied to any other CDO tranche — ELM can be applied to any set of cash flows generated by the same CDO.

A Moody’s rating is typically assigned to both an annual rate of return and the return of the initial principal investment. The most common Moody’s rating assumes an annual rate of return equal to a security’s stated coupon.59 For example, if a plain vanilla CDO security has a stated coupon of 8% and the targeted rating for the security is Baa2, Moody’s would discount the cash flows to that security at 8% and the rating would address the return of the initial principal investment at that discount rate.

57 The distribution of the loss scenarios are modeled under the Moody’s Binomial Expansion Technique with the number of scenarios equal to the Diversity Score (D) of the CDO collateral assets. For more information, please refer to The Binomial Expansion Method Applied to CBO/CLO Analysis, Moody’s Investors Service, December 1996.

58 Expressed as a function of rating and tenor based on the historical loss experience of Moody’s-rated securities.

59 In other words, the entire stated coupon and principal of such security will be encompassed by the Moody’s rating.

The lower the expected loss for a security, the

higher its rating.

The same Moody’s methodology is

applied to all CDO securities, including

combination securities.



With respect to a combination security, however, a Moody’s rating may address an annual rate of return that is less than its stated coupon rate (e.g., a bond with an 8% coupon where only a 2% of the coupon and the principal amount are rated Baa2. When the ELM is applied to this CDO security, Moody’s would use a 2% discount rate (not 8%) when calculating the net present value of the cash flows. Since this rating is derived using a discount rate lower than 8%, the expected loss calculated is lower and this implies a rating higher than Baa2 for the security.60

Finally, ELM can be extended to CDO combination securities (which typically do not have a stated coupon rate) and any stream of cash flows generated by a CDO.61 The investor determines the minimum rate of return that the rating addresses (the discount rate used for calculating the net present value of the cash flows). As a rule of thumb, for any given security or a stream of cash flows, the higher the desired minimum rated return, the higher the expected loss for the security and the lower its rating.

Using the investment grade CDO in Figure 99, assume a combination security is created using class B notes (62%) and equity (38%). In order to analyze the rating of this combination security under two different rated coupons (0%62 and 4%), the same ELM would be applied in both cases, but two different discount rates would be used to calculate the NPVs. One would expect the rating that addresses a lower coupon to have a lower expected loss and a higher rating. That is the case: the combination security with the 0% rated coupon garners a rating of A2, while the security with the 4% rated coupon reaches only a Baa2 rating.

Figure 99. Investment Grade Average CDO Example (Dollars in Millions) Size (in $MM) Coupon (%) Rating WAL Class A-1 407.5 Aaa 8.3Class A-2 39.0 6.66 Aa2 10.5Class B 33.5 8.00 Baa2 11.1Equity 22.0 — NR —

Note: The average rating of the portfolio is Baa3, and the legal final on the securities is 12 years. A portion of Class A-2 and B can be sold on a floating-rate basis. Source: Salomon Smith Barney.

Combination Securities With the Same Rating Are Not Created Equal For any given CDO, a variety of similarly rated combination securities (with the same expected loss) using different parts of the underlying CDO capital structure can be created. As Figure 100 illustrates, we have created six Baa2-rated combination securities from the classes in the investment grade-average CDO in Figure 99.

60 By applying a lower discount rate, the net present value of the cash flows would be higher in any given scenario, which would result in a lower loss number.

61 A CDO combination security usually does not have a stated expected coupon because one of its components is the equity tranche. The equity tranche is entitled to receive excess cash flows and not an expected coupon rate (this differs from the rated coupon, the rate of return covered by the credit rating).

62 A rating that addresses a coupon of 0% is sometimes called a “principal-only rating.”

ELM can be extended to CDO combination

securities and any stream of cash flows generated by a CDO.

Generally, the higher the desired minimum

rated return, the higher the expected loss for

the security and the lower its rating.



Figure 100. Different Mixes of Baa2-Rated Combination Securities Components Rating Rated Coupon (%) Class A-2 (%) Class B (%) Equity (%) C1 Baa2 0 30 70C2 Baa2 0 35 65C3 Baa2 2 42 58C4 Baa2 2 48 52C5 Baa2 4 58 42C6 Baa2 4 62 38


Despite the fact that all six securities have the same Baa2 rating, their other characteristics differ significantly, as shown in Figure 101. Their internal rates of return are different, and the effects of changes in constant default rates on those returns also differ.63 For example, the expected IRR for a constant default rate of 0.25% per annum, given rated coupons of 0% is roughly 14.7%, whereas the IRR for 4%-rated coupons is roughly 11.5%. Also, note that the range of returns under various default scenarios is about 16% for securities with 0% coupons, but only 8% for the 4% coupons.

Figure 101. Performance of Different Mixes of Baa2-Rated Combination Securities (the Graph Shows the IRR as a Function of Default Rate with the Raw Data Shown)

-2%

2%

6%

10%

14%

18%

0.00% 0.25% 0.50% 0.75% 1.00% 1.25% 1.50% 1.75% 2.25%Annual Default Rate

IRR

C1 C2 C3

C4 C5 C6

C1 C2 C3 C4 C5 C6 Rating Baa2 Baa2 Baa2 Baa2 Baa2 Baa2 Rated Coupon 0% 0% 2% 2% 4% 4% Annual Default Rate Internal Rate of Return 0.00% 16.0% 15.7% 14.2% 14.1% 12.2% 12.3%0.25% 14.7 14.6 13.1 13.2 11.3 11.60.50% 13.3 13.3 11.9 12.1 10.4 10.80.75% 11.8 11.9 10.6 10.9 9.4 10.01.00% 9.9 10.2 9.1 9.6 8.3 9.11.25% 7.5 8.2 7.3 8.1 7.1 8.11.50% 5.0 6.0 5.4 6.5 5.8 7.11.75% 2.2 3.7 3.5 4.9 4.6 6.12.25% -0.8 0.7 1.5 2.7 3.4 4.4


63 In addition to the rating, the expected loss threshold of a particular CDO security is also based on the expected tenor of the security (the shorter the tenor, the lower the expected loss). For the purposes of this report, we have simplified the analysis by assuming that all securities have the same tenor.

The six securities in Figure 100 are similar

in rating only.



Also, there is a relationship between the rated coupon on a combination security and the mix of CDO tranches that support that combination security. That is, for a given rating and the same set of components, if a higher-rated coupon is desired, a higher proportion of the senior tranche is needed. For example, C1 and C5 are both made of the same components, but in different proportions because C5 needs a higher proportion of the “safer” Class A-2 security (which reduces expected loss) to compensate for a higher-rated coupon or discount rate (which reduces the NPV of the cash flow stream, thereby generating a higher expected loss).

In instances where the combination security rating is in the low investment grade area (a higher tolerable amount of expected losses relative to high investment grade), however, the loss contribution of the rated components to the overall expected loss number of the combination security is quite insignificant. The reason is that losses on the rated components are discounted using a rated coupon that is lower than their stated coupon rate, so the portion of the expected losses corresponding to the rated components are lower than their actual ratings would imply.64 For this reason, the Class B Note can “afford” to combine itself with a relatively similar amount of equity as the Class A-2 Notes and still get the same Baa2 rating on the combination securities, given the same-rated coupon.

Monitored Versus Nonmonitored Ratings When Moody’s issues a rating for a given combination security, it is either a monitored rating or nonmonitored rating. A nonmonitored rating is a rating that is issued at a given point in time (usually on the closing date of the underlying CDO) with no subsequent surveillance or monitoring work with respect to such rating. The rating is not transferable when the combination security is traded because it only addresses the transaction at issuance. A monitored rating (as the name implies), is a rating that is kept current by Moody’s, which reviews it for any possible changes in creditworthiness. Almost every plain vanilla Moody’s rating falls into the latter category.

One key consideration for a monitored rating is the amortization methodology applied to the combination security: what principal amount is outstanding and monitored? For example, with respect to a combination security with a principal-only rating (i.e., rated coupon of 0%), each cash flow stream received from each of the underlying components is a return of principal from the rating perspective. Every dollar is used to amortize the “rated principal balance.” Conversely, assume that a $100 million Baa2-rated combination security with a minimum-rated coupon of 4% receives $10 million of cash flow on its first semiannual payment date. Since the rated coupon is 4% per annum (or 2% on any semiannual payment date) on the rated principal balance, $2 million of the $10 million would be treated as “rated coupon,” and the other $8 million would be treated as return of “rated principal.” The rating of the security would address an amortized principal balance of $92 million after the first payment date until another payment in excess of the rated coupon (calculated on the rated principal balance) is received. For a given monitored rating, the lower the rated coupon, the faster the rated balance will decline.

64 For example, an 8% Baa2-rated bond derived its expected loss by discounting its cash flows using 8%. If the same bond/cash flows are discounted using a lower rate (0% for example), the expected loss would be significantly lower (i.e., a higher rating).

For a given rating and same set of components,

the higher the rated coupon on the

combination security, the higher the proportion

of senior tranche.

How are CDO combination

securities amortized and monitored?



Alternatively, the rated principal balance can increase if the cash flows received on any payment period are insufficient to cover the rated coupon. The rated principal balance concept mirrors that of a pay-in-kind security where (i) the amount of cash flow shortfall to the rated coupon would be capitalized and treated as additional rated principal balance, and (ii) the rated principal balance is used to determine the rated coupon.

Since the expected return of a combination security is always higher than its rated coupon, it is possible that a monitored rating would be withdrawn when the rated balance reaches zero at the tail end of a CDO transaction, even though the combination security is still expected to pay out cash flows from that point on.65

65 In practice, this is not a problem, since combination securities can be easily modified to prevent such effects.



Return Analysis

Combination Security Pricing — Current Market Convention As we have mentioned, the conventional CDO combination security analytical methodology is limited because the security’s rating, rated coupon and expected yield gives no information to an investor regarding the dispersion or variability of returns around the mean return. As Figure 101 illustrates, two combination securities can have the same rating and rated coupon, but have highly different return profiles. The constant default rate method is limited in this respect.

Under the constant default rate method, multiple default and recovery rate scenarios are used to compute a risk-adjusted IRR on its initial capital outlay. The standard approach includes a scenario that assumes a constant default rate (CDR) of 2% per annum for high-yield CDOs and 0.25% for investment-grade CDOs.66 In addition, various default “spike” scenarios are run. Default rate spikes are assumed to occur at the beginning, middle, and end of the transaction’s life. Under each scenario, an IRR is calculated from the resulting cash flows, and it is used as the expected value of the return.

Market convention notwithstanding, there are certain weaknesses in analyzing CDO combination securities based on a constant default rate methodology. First, annual default rates are not uniformly distributed in either investment-grade or high-yield rating categories. To illustrate this point, Figure 102 shows frequency distributions of annual default rates for bonds rated Baa3 and B2.67 For Baa3 bonds (left panel), the mean is 0.31%,68 close to the 0.25% constant default rate assumption commonly used to price investment-grade CDOs.69 The frequency distribution of Baa3 default rates is highly skewed, however, with the closest approximation being either a Poisson or negative exponential distribution. Also, notice that the standard deviation of Baa3 credits is 0.9%, but that at -1 standard deviation Baa3 credits have an annual default rate of negative 0.59%. Similarly, for B2 credits (right panel), the distribution is highly skewed, such that the distribution is more Poisson than normal, and the standard deviation is inadequate as a measure of dispersion.

66 This target scenario usually assumes a recovery rate of 40%–50%.

67 The average rating of a typical investment-grade CDO portfolio is Baa3, and the average rating of a typical high-yield CDO portfolio is B2.

68 Please note that this 1970–2000 average one-year default rate (i.e., 0.31%) differs from the approximately 0.5% average one-year default rate we reported for the period 1983–2000 earlier in this paper because from 1970–1983, Moody’s only published Baa default rates.

69 Default and Recovery Rates of Corporate Bond Issuers: 2000, Exhibits 39 and 40, Moody’s Investors Service, February 2001. Because Moody’s does not report ratings data by modifier before 1983 (i.e., Baa1, Baa2 and Baa3), Baa one-year default rates were used for the 13 years from 1970 to 1982. From 1983 to 2000, the average one-year default rate for Baa3 credits is 0.46%.

A security’s rating, rated coupon, and

expected yield give an investor no information

about the potential volatility of the return.

Since corporate default rates are not uniformly

distributed, the constant default rate method is

inherently flawed.



Figure 102. Annual Default Rates from Moody’s for Bonds Rated Baa3 (Left) and B2 (RIght), 1970–2000

0

5

10

15

20

25

0.13% 0.88% 1.63% 2.38% 3.13% 3.88% 4.63% 5.38%Annual Default Rate

Freq

uenc

y Baa3Mean = 0.31%Median = 0.0%

Std. Dev. = 0.9%

0

1

2

3

4

5

6

0.50% 4.50% 8.50% 12.50% 16.50% 20.50%Annual Default Rate

Freq

uenc

y

B2Mean = 7.09%

Median = 5.41%Std.Dev. = 5.9%

a Default rates prior to 1983 are available by letter category only. Sources: Moody’s Investor Service and Salomon Smith Barney.

If the mean is not a good measure of central tendency for triple-B minus and single-B credits, can the median value serve as a reliable benchmark for pricing CDO combination securities? No. The median of the triple-B minus distribution is zero, yet no rational person would argue that there is no default risk for investment-grade CDO combination security investors. Likewise, even though the mean and median are similar for B2 default rates, large annual default rate spikes occur more than occasionally.

Finally, with respect to the timing of defaults, both investment-grade and high-yield CDO combination security returns are critically path dependent. That is, large defaults that occur early in the life of the CDO have a much greater negative effect on returns than later defaults. Timing is key. Thus, even one early year with a 10% default rate for the high-yield CDO or an early 5% annual default rate for an investment-grade CDO can impair the return on CDO combination securities.

Pricing Combination Securities Using Monte Carlo Simulations In this section, we propose a refinement to the current market convention for pricing CDO combination securities — a refinement based on Monte Carlo simulations and actual historical default rates. The method is a generalization of that which we proposed earlier for analyzing investment grade CDOs.70 We use the capital structure and initial portfolio from a generic investment-grade average CDO (Figure 99).

The fact that the distribution of annual default rates is not normal, and is not described well by traditional nonnormal distributions, presents difficulties in generating meaningful annual default paths for the simulation. Our approach is to make no parametric assumptions about the underlying distribution, but rather to generate annual default rates for each pricing vector by sampling (with replacement) from the historical default rates published by Moody’s since 1970.71 Accordingly, we randomly sampled from these annual default rates and constructed 2,000 default vectors that are a mix of front-, middle-, and back-loaded curves (see Figure 103 for a sample of paths). For example, Figure 103 shows a sample of default rates along eight paths used for CDO pricing.

70 McDermott, G., Benzschawel, T. and Khan, S. Investment-Grade CDOs, Salomon Smith Barney, August, 2001.

71 Default and Recovery Rates of Corporate Bond Issuers: 2000, Moody’s Investors Service, February 2001. We assumed a typical credit distribution from an investment-grade average CDO. The distribution, which we call the “tight” distribution, can be found in Figure 21 of Investment Grade CDOs.

Default timing is critical for CDO payouts.

Our method is grounded in historical

default rates and Monte Carlo simulations.

We randomly sampled historical annual default

rates and constructed 2,000 default vectors that are a mix of front, middle, and back-loaded curves.



Figure 103. Sample of Eight Paths Used to Price a 15-Year Investment-Grade CDO

0%

1%

2%

3%

4%

5%

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15Time (Years)

Annu

al D

efau

lt Ra

te

Sources: Moody’s Investors Service and Salomon Smith Barney.

We then took the capital structure and initial portfolio of a recent investment-grade average CDO and computed the cash flows along the 2,000 simulated default vectors. For the purposes of calculating ex-ante Sharpe ratios,72 cash flows for a given tranche were discounted back to the pricing date, using values obtained from the yield curve on that date.73 We then calculated the mean, standard deviation, and Sharpe ratio on the combination security cash flows.74

72 See Sharpe, W.F. “The Sharpe Ratio,” Journal of Portfolio Management, 21, 1994, pp 49–54.

73 January 7, 2002.

74 The Sharpe ratio is calculated as the incremental return on an investment over that of a risk-free asset for a given regular sampling divided by the standard deviation of the excess returns over the sample period. It is customary to express the Sharpe ratio as an annualized number. For example, Sharpe states that typical estimates of annual excess return on the stock market of a developed country would be a Sharpe Ratio of 0.4 (see footnote on Securities Investor Protection Corporation). For this particular case, the risk/reward ratio is called the “ex-ante Sharpe ratio” to indicate that it is an expected value.



Application of SSB MC Methodology to Combination Securities Figure 104 compares returns on various combination securities using the conventional CDR pricing methodology and Salomon Smith Barney Monte Carlo Pricing Methodology. The five leftmost columns of the figure illustrate key features of various combination structures. The first column on the left shows the minimum rating for the combination grouped by rated coupon (except for the top three rows that are the plain vanilla notes and pure equity).75 The third, fourth, and fifth columns provide information on the capital structure of the securities in terms of percentages of senior, mezzanine, and equity tranche contributions, respectively.

75 There is no minimum coupon for the plain vanilla single-tranche assets.



Figure 104. Conventional CDR Pricing Method Versus SSB Monte Carlo Default Simulation Method for Combination Securities Indicative Data Constant Default Monte Carlo Simulation

Rated Combinations (%) (0.25% Per Annum) Sharpe Internal Rate of Return (%)

Rating Coupon (%) Class A-2 Class B Equity IRR (%) Mean IRR (%) StdDev of IRR RatioMin Q1 Median Q2 Max Baa2 8.00 0 100 0 8.0 8.0 0.0 95.0 8.0 8.0 8.0 8.0 8.0Aa2 6.66 100 0 0 6.7 6.7 0.0 1601.5 6.7 6.7 6.7 6.7 6.7Equity NA 0 0 100 18.8 18.2 1.4 1.7 8.3 17.7 18.5 19.1 20.3

Baa2 0 30 0 70 14.7 14.2 1.1 1.8 7.6 13.8 14.4 14.9 15.8Baa2 0 0 35 65 14.6 14.1 1.0 1.9 8.2 13.7 14.3 14.7 15.5Baa2 2 42 0 58 13.1 12.7 0.9 1.8 7.4 12.3 12.9 13.3 14.1Baa2 2 0 48 52 13.2 12.8 0.8 2.1 8.1 12.5 13.0 13.3 14.0Baa2 4 58 0 42 11.3 11.0 0.7 2.0 7.1 10.7 11.1 11.5 12.0Baa2 4 0 62 38 11.6 11.3 0.6 2.4 8.1 11.0 11.4 11.7 12.2A2 0 48 0 52 12.5 12.1 0.8 1.9 7.3 11.8 12.3 12.7 13.4A2 0 0 58 42 12.1 11.8 0.7 2.3 8.1 11.5 11.9 12.2 12.8A2 2 60 0 40 11.0 10.7 0.6 2.0 7.1 10.4 10.9 11.2 11.7A2 2 0 70 30 10.8 10.6 0.5 2.7 8.1 10.4 10.7 10.9 11.3A2 4 75 0 25 9.3 9.1 0.4 2.3 6.9 8.9 9.2 9.4 9.8A2 4 0 85 15 9.4 9.3 0.2 4.2 8.0 9.1 9.3 9.4 9.6Aa2 0 60 0 40 11.0 10.7 0.6 2.0 7.1 10.4 10.9 11.2 11.7Aa2 0 43 45 12 8.6 8.5 0.2 4.0 7.4 8.4 8.5 8.6 8.8Aa2 2 70 0 30 9.9 9.6 0.5 2.2 7.0 9.4 9.8 10.0 10.4Aa2 2 55 35 10 8.2 8.1 0.2 4.4 7.3 8.0 8.1 8.2 8.4Aa2 4 82 0 18 8.5 8.4 0.3 2.7 6.8 8.2 8.4 8.6 8.8Aa2 4 70 22 8 7.7 7.7 0.1 5.0 7.1 7.6 7.7 7.8 7.9




The column labeled Constant Default in Figure 104 provides the IRR for each of the combinations specified when priced with a constant annual default rate of 0.25%.76 By contrast, the next eight columns provide summary data from our Monte Carlo simulations. The mean IRR from the simulations appear in the leftmost column of the simulation data followed by the standard deviation of those IRR over the 2000 default paths. We also calculated ex-ante Sharpe ratios.77 The IRRs at the quartiles of the return distribution (Q1, Median, and Q3) and the minimum and maximum IRRs over the 2000 paths for each deal are also shown.

With respect to the conventional CDR methodology, the largest IRR in the table (18.8%) is obtained for the pure equity investment. This is also true for the pure equity investment analyzed using the Monte Carlo methodology (18.2%). In fact, an examination of IRR calculated under both pricing methodologies suggests strongly that the size of returns is directly related to the amount of equity backing the combination security. This is confirmed by the histograms in Figure 105, which plot the IRR obtained under the constant 0.25% per annum default assumption (dark bars) and using the Monte Carlo simulator (light bars). Figure 105 clearly illustrates that returns are almost entirely dependent on the percentage of equity in the capital structure. Since the Baa/Baa3 average default rate over the 1970–2000 period is 0.31%,78 one would expect the average IRR obtained from the different pricing methods to be nearly identical.79 In fact, the proximity between average IRR obtained under the two methods supports the notion that our Monte Carlo simulator is operating correctly.

Figure 105. Percentage of Equity in a Combination Security and Its Impact on Returns

5%

10%

15%

20%

0% 8% 13% 18% 30% 38% 40% 43% 53% 65% 100%Percentage of Equity

IRR

Constant Default

Monte Carlo


76 The 0.25% constant default rate was used for pricing as it is close to the average annual default rate for investment-grade bonds and is commonly assumed when pricing investment-grade CDOs using the constant default method.

77 Ex-Ante Sharpe ratios were calculated by dividing the average excess return (in dollars) by the standard deviation of the excess returns over the 2000 default paths and annualizing that number by dividing by the square root of the life of the deal (in years).

78 Recall from footnote on Securities Investor Protection Corporation (SIPC) © Citigroup. Because Moody’s does not report ratings data by modifier before 1983 (i.e. Baa1, Baa2, Baa3), Baa2 one-year default rates were used for the 13 years from 1970 to 1982.

79 The small differences among IRRs for the two calculation methods is due to the fact that the historical average of the default rates is only slightly more than the 0.25% assumption used for the constant default calculations.

The combination notes proved to be a very

stable asset class, with Sharpe ratios ranging

from 1.8 to 5.0.

The absolute level of returns is directly

proportional to the amount of equity

backing the combination security.



One interpretation of Figure 105 is that a yield-driven investor might be wise to shun combination securities altogether in order to purchase an all-equity tranche. Information solely about average returns, however, provides an investor with little information regarding the volatility of those returns. For example, even though the expected IRR for the pure equity tranche is the highest, the distribution of expected returns could be very wide, exposing an investor to potentially high fluctuations in annual profits and losses. It is impossible to accurately quantify this type of risk using the conventional CDR method. This is because the cash flows from a CDO are highly dependent on the temporal pattern of default as well as the overall average rate. This temporal dependency can be captured using Monte Carlo simulations and the effects of even greater-than-historical stresses can be performed using a multiplier of historical rates in the simulations.

The rightmost seven columns in Figure 104 illustrate the richness of the information provided by the default simulation method. For example, the standard deviations of the IRRs obtained from the 2000 simulation paths used to price each combination security are shown along with the resulting Sharpe ratio for that security. The extremely high Sharpe ratios for the plain vanilla Baa2 and Aa2 tranches indicate that they are nearly riskless under pricing simulations grounded in historical default rates.80 The pure equity investment, by contrast, has more risk, but its Sharpe Ratio of 1.7 is still impressive (see Footnote 74). The combination security with the highest Sharpe Ratio (5.0) is rated Aa2 with a minimum coupon of 4% and a ratio of equity to mezzanine to senior components of 8:22:70. Its expected IRR is 7.7%. In other words, although the pure equity investment has the highest expected return at 18.2%, its Sharpe ratio (1.7) is about one-third that of the Aa2/4%/8:22:70 combination security. In fact, the Sharpe ratios for all the combination securities are greater than that for the pure equity tranche.

The subsequent columns of Figure 104 display the minimum IRR, the IRR at the quartiles and the maximum IRR. Those values indicate that the IRRs along all pure equity pricing paths are positive, but highly variable, returning between 8.3% and 20.3%, whereas the IRRs for Aa2/4%/8:22:70 combination security range between 7.1% and 7.9%.

One striking feature of all the structures in Figure 104 is that none of the paths have negative returns (even though cash flows are missed along some of the pricing paths). Thus, as we demonstrated in our earlier work (see Footnote 80), investment-grade CDO equity is an attractive asset class by itself. However, not all investors wish to take on the volatility of returns implied from the simulations in Figure 104 to obtain the 18.2% expected return on their investment. Some investors are likely to be willing to give up some expected yield for less volatile earnings (i.e., bracketed yield target on their return). Also, some investors have guidelines that require a minimum rating and/or a rated coupon.

80 This result is pointed out in our initial simulations reported in McDermott, G., Benzschawel, T. and Khan, S. Investment-Grade CDOs, Salomon Smith Barney, August, 2001.

The pure equity investment had the

highest IRR, but not the highest Sharpe ratio.

Some investors are likely to be willing to

give up some expected yield for less volatile earnings, and some have guidelines that

require a minimum rating and/or a

minimum-rated coupon.

All pricing paths for the equity tranche show

positive returns, but the range between

minimum and maximum values is much greater

than that of the combination securities.



Figure 106 is a plot of IRR (open symbols referenced to left axis) and Sharpe ratios (solid symbols referenced to right axis) for the various combination structures in Figure 104 ordered by percentage of equity from highest to lowest. The solid gray lines in the figure trace the minimum and maximum IRR obtained over the 2000 default paths used in the pricing each security. In Figure 106 each deal is represented as a combination of rating, rated coupon, and proportion of equity, mezzanine, and senior tranches. The figure shows that, in general, as the proportion of equity in the deal goes up, the expected IRR increases and the Sharpe ratio declines.

Figure 106. Expected Yield/Sharpe Ratio Trade-Offs for Various Combination Securities, Ordered by Percentage of Equity

5%

10%

15%

20%

25%

Equit

y/NA/10

0:0:0

Baa2/0

%/70:0:

30

Baa2/0

%/65:35

:0

Baa2/2

%/58:0:

42

Baa2/2

%/52:48

:0

A2/0%/52

:0:48

A2/0%/42

:58:0

Baa2/4

%/42:58

:0

A2/2%/40

:0:60

Aa2/0%

/40:0:

60

Baa2/4

%/38:62

:0

A2/2%/30

:70:0

Aa2/2%

/30:0:

70

A2/4%/25

:0:75

Aa2/4%

/18:0:

82

A2/4%/15

:85:0

Aa2/0%

/12:45

:43

Aa2/2%

/10:35

:55

Aa2/4%

/8:22

:70

Baa2/N

A/0:10

0:0

Aa2/NA/0:

0:100

Rating / Coupon / Equity:Mezzanine:Senior

IRR

1.5

2.5

3.5

4.5

5.5100% 70% 65% 58% 52% 52% 42% 42% 40% 40% 38% 30% 30% 25% 18% 15% 12% 10% 8% 0% 0%

Percentage of Equity in Combination Security

Shar

pe ra

tio

IRRMin IRR and Max IRRSharpe Ratio

A brief example to clarify the abbreviations on the horizontal axis: Baa2/4%/50:50:0 refers to a Baa2 rated combination security with a minimum rated coupon of 4% that is backed by 50% equity, 50% mezzanine tranche and 0% subordinated tranche. Source: Salomon Smith Barney.

One important feature of the results in Figure 106 is that, for a given deal, the deviations of the returns along different pricing paths from the average IRR of the deal are not symmetric. That is, higher-than-average IRRs are generally closer to the average return than are below-average returns, and median returns are generally higher than mean returns. It might make more sense for an investor to move across Figure 106 from left to right and identify an investment that has a much tighter relationship between minimum and maximum returns and a higher Sharpe ratio. This is an important characteristic of the return profile from such deals, and one that is not easily revealed using the constant default pricing method.

Consider how an investor might use the information contained in Figure 106 to optimize their investment objectives. Assume that a fund manager must invest in only A-rated and above securities, and that management will tolerate only a 3% deviation from the average expected annual return on any investment over $50 million. Examination of Figure 106 indicates that all but one combination security to the right and including that labeled A2/2%/40:0:60 (see arrow in Figure 106) will satisfy the A-level rating criterion (as well a couple of deals to the left). However, all but one security at and to the right of A2/2%/30:70:0 will satisfy the 3% IRR deviation criterion.81 By investing in the A2/2%/30:70:0 combination security, the 81 One can use Global Markets Inc., 2004 to determine if a given deal meets the 3% range criterion.

As the proportion of equity in the deal goes

up, the expected IRR increases and the

Sharpe ratio declines.

Deviations of returns from the average IRRs

are not symmetric.



investor would achieve an expected IRR of over 10% with a Sharpe ratio of 2.7, with the knowledge that, under historical conditions, that IRR would be bounded at 7.0% and 10.4%.

SSB Monte Carlo Default Simulation Methodology — Stressed Case One limitation of our Monte Carlo simulation method (which is grounded in historical default rate data) is that future default rates may be greater than those realized over the past 30 years. Because we recognize this possibility, and because we wish to examine the effect of credit stress on our investment-grade CDOs and combination securities, we ran our pricing simulations using a 2x multiple on the historical default rates. The results of those simulations, plotted in a similar form to Figure 106, appear in Figure 107 and in tabular form in Appendix I.

The figure shows that, in general and as in Figure 106, as the proportion of equity in the deal goes up, the expected IRR increases and the Sharpe ratio declines. As in Figure 106, if a return is below the mean, there is the risk that the magnitude below the mean is much greater than the magnitude of returns above the mean. In contrast to Figure 106, the data in Figure 107 show that stressing default rates by 2x produces pricing paths that have negative returns. This occurs for the pure equity tranche and for the Baa2/0%/70:0:30 combination security. Thus, the results in Figure 107 indicate that, not surprisingly, tranches containing greater amounts of equity are more susceptible to higher default rates.

Figure 107. Expected Yield/Sharpe Ratio Trade-Offs for Various Combination Securities Using 2x Historical Default Rates, Ordered by Percentage of Equity

-30%

-20%

-10%

0%

10%

20%

Equit

y/NA/100:0:0

Baa2/0%/70:0:30

Baa2/0%/65:35:0

Baa2/2%/58:0:42

Baa2/2%/52:48:0

A2/0%/52:0:48

A2/0%/42:58:0

Baa2/4%/42:58:0

A2/2%/40:0:60

Aa2/0%/40:0:60

Baa2/4%/38:62:0

A2/2%/30:70:0

Aa2/2%/30:0:70

A2/4%/25:0:75

Aa2/4%/18:0:82

A2/4%/15:85:0

Aa2/0%/12:45:43

Aa2/2%/10:35:55

Aa2/4%/8:22:70

Baa2/NA/0:100:0

Aa2/NA/0:0:100

Rating / Coupon / Equity:Mezzanine:Senior

IRR

0.0

1.0

2.0

100% 70% 65% 58% 52% 52% 42% 42% 40% 40% 38% 30% 30% 25% 18% 15% 12% 10% 8% 0% 0%

Percentage of Equity in Combination Security

Shar

pe ra

tioIRR

Min IRR and Max IRRSharpe Ratio


We ran our analysis assuming a 2x multiple

of historic default rates.

Even under the extremely stressful and

unrealistic 2x stress, returns on combination securities held up well.



How might the information in Figure 107 affect our hypothetical fund manager’s investment decision? Investment in the A2/2%/30:70:0 combination security, whose IRR and Sharpe ratio under historical default conditions is 10.6% and 2.7, respectively, would no longer provide the best trade-off between IRR (9.7%) and Sharpe ratio (1.1) under stressed conditions. Since the A2/2%/30:70:0 combination security may be adversely affected under the onerous 2x stress, the manager may wish to invest in the less risky combination note. For example, A2/4%/15:85:0 combination security with an expected IRR of 9.3% and a Sharpe ratio of about 4.2 under historical default conditions, and an expected IRR of 8.8% and Sharpe ratio of 1.9 under stressed conditions, would be a good candidate. It would satisfy all investor constraints in that its 2x-stressed minimum IRR is 6.5%, within the 3% deviation criterion from its expected return of 9.3% under historical conditions.

Methodology Limitations It is important to note the limitations of the methods employed for this analysis and the extent to which they fail to provide complete assessments of the risks involved in investing in CDO combination securities. The main limitations of our analytical method are: (1) the cost of lack of liquidity to CDO investors is not assessed; (2) the effect of active portfolio management on returns is not quantified; (3) a constant 50% recovery rate is assumed; (4) no temporal correlation of annual default rates is assumed; (5) the 30-year sample of default rates may not be sufficient to characterize the future distribution of rates; (6) we assume that the manager maintains a portfolio with an average credit quality of Baa3; and (7) there has been no attempt to analyze each deal at the individual asset level.



Conclusion

CDO combination securities can be tailored for each investor based on the desired rating, rated coupon, yield target, and capital guidelines. We examined how varying combinations along those dimensions affect IRRs and Sharpe ratios using both traditional and simulation-based analytics. We find that securities backed by investment grade bonds produce high expected returns and Sharpe ratios under historical default conditions. Also, we find a direct relationship between expected IRR and the proportion of CDO equity in each security, but an inverse relationship between CDO equity and Sharpe ratio. Not all similarly rated combination securities produce equal expected returns and may differ greatly in Sharpe ratios. The results of our analyses can be used by investors to customize the mix of CDO senior, mezzanine, and equity tranches to achieve their investment objective.



Appendix

In this section, we present results of stress testing returns using both the constant default method and the Salomon Smith Barney Monte Carlo Simulation Method.

Figure 108 shows IRRs computed for the combination structures having various ratings, rated coupons, and capital structures using constant default rates ranging from 0.0% to 2.25% in steps of 0.25%. Recall that the historical average default rate for investment-grade corporate bonds is roughly 0.25%.

Figure 109 shows IRRs computed for the combination structures using the Monte Carlo simulation method with a 2x multiple on every point in the historical data sample of one-year defaults used in the simulation.

Figure 110 illustrates the computation of the historical sample used in Monte Carlo simulation. Each of the 30 data points was calculated as a weighted average of the actual one-year default rates of A2- to Ba3-rated bonds from 1970 through 2000. The weight for each rating corresponds to the percentage of bonds with that rating in the CDO’s collateral pool.



Figure 108. Constant Default Pricing of Various Combination Securities Using Constant Default Rates Above and Below the Historical Average of 0.25% Rated Combinations (%) IRR ( Constant Default) (%)

Rating Coupon (%) Class A-2 Class B Equity 0.00% 0.25% 0.50% 0.75% 1.00% 1.25% 1.50% 1.75% 2.00% 2.25%Baa2 8.00 0 100 0 8.0 8.0 8.0 8.0 8.0 8.0 8.0 8.0 8.0 7.3Aa2 6.66 100 0 0 6.7 6.7 6.7 6.7 6.7 6.7 6.7 6.7 6.7 6.7Equity NA 0 0 100 20.4 18.8 17.1 15.0 12.4 8.3 3.1 -4.2 -21.4 -28.1

Baa2 0 30 0 70 16.0 14.7 13.3 11.8 9.9 7.5 5.0 2.2 -0.1 -0.8Baa2 0 0 35 65 15.7 14.6 13.3 11.9 10.2 8.2 6.0 3.7 1.9 0.7Baa2 2 42 0 58 14.2 13.1 11.9 10.6 9.1 7.3 5.4 3.5 2.0 1.5Baa2 2 0 48 52 14.1 13.2 12.1 10.9 9.6 8.1 6.5 4.9 3.8 2.7Baa2 4 58 0 42 12.2 11.3 10.4 9.4 8.3 7.1 5.8 4.6 3.7 3.4Baa2 4 0 62 38 12.3 11.6 10.8 10.0 9.1 8.1 7.1 6.1 5.4 4.4A2 0 48 0 52 13.5 12.5 11.4 10.2 8.8 7.2 5.6 3.9 2.6 2.2A2 0 0 58 42 12.9 12.1 11.2 10.3 9.2 8.1 6.9 5.7 4.9 3.9A2 2 60 0 40 11.8 11.0 10.2 9.2 8.2 7.1 5.9 4.8 4.0 3.7A2 2 0 70 30 11.4 10.8 10.2 9.5 8.8 8.1 7.3 6.6 6.0 5.1A2 4 75 0 25 9.9 9.3 8.8 8.2 7.5 6.9 6.2 5.6 5.2 5.0A2 4 0 85 15 9.7 9.4 9.0 8.7 8.4 8.0 7.7 7.4 7.1 6.3Aa2 0 60 0 40 11.8 11.0 10.2 9.2 8.2 7.1 5.9 4.8 4.0 3.7Aa2 0 43 45 12 8.9 8.6 8.3 8.0 7.7 7.4 7.1 6.9 6.7 6.2Aa2 2 70 0 30 10.5 9.9 9.2 8.5 7.7 6.9 6.1 5.4 4.8 4.6Aa2 2 55 35 10 8.4 8.2 7.9 7.7 7.5 7.2 7.0 6.8 6.7 6.3Aa2 4 82 0 18 8.9 8.5 8.1 7.7 7.2 6.8 6.4 6.0 5.7 5.6Aa2 4 70 22 8 7.9 7.7 7.6 7.4 7.2 7.0 6.9 6.7 6.6 6.4




Figure 109. Stress Testing Various Combination Securities using Monte Carlo Simulations and 2x Historical Default Rates Rated Combinations (%) Monte Carlo (2x Default Rates)

Rating Coupon Class A-2 Class B Equity Mean IRR (%) Sharpe Min IRR (%) Max IRR (%)Baa2 8.00 0 100 0 8.0 7.2 8.0Aa2 6.66 100 0 0 6.7 6.7 6.7Equity NA 0 0 100 15.0 0.61 -32.7 20.0

Baa2 0 30 0 70 12.0 0.66 -1.3 15.6Baa2 0 0 35 65 12.1 0.73 0.6 15.4Baa2 2 42 0 58 10.9 0.69 1.1 13.9Baa2 2 0 48 52 11.2 0.82 2.7 13.8A2 0 48 0 52 10.4 0.71 1.9 13.2Baa2 4 58 0 43 9.6 0.76 3.2 11.9A2 0 0 58 43 10.5 0.92 3.9 12.7A2 2 60 0 40 9.4 0.77 3.5 11.6Aa2 0 60 0 40 9.4 0.77 3.5 11.6Baa2 4 0 62 38 10.2 0.99 4.5 12.1A2 2 0 70 30 9.7 1.14 5.2 11.2Aa2 2 70 0 30 8.7 0.86 4.4 10.3A2 4 75 0 25 8.3 0.94 4.9 9.7Aa2 4 82 0 18 7.8 1.12 5.5 8.8A2 4 0 85 15 8.8 1.92 6.5 9.6Aa2 0 43 45 12 8.1 1.81 6.3 8.8Aa2 2 55 35 10 7.8 2.02 6.3 8.3Aa2 4 70 22 8 7.4 2.32 6.4 7.9


Figure 110. Sample of Annual Default Rates Used in the Monte Carlo Simulation Weights 2.5% 5.0% 12.5% 15.0% 50.0% 7.0% 5.0% 3.0% SampleRating A2 A3 Baa1 Baa2 Baa3 Ba1 Ba2 Ba3Year 1970 0.00% 0.00% 0.27% 0.27% 0.27% 4.12% 4.12% 4.12% 0.83% 1971 0.00 0.00 0.00 0.00 0.00 0.42 0.42 0.42 0.06 1972 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1973 0.00 0.00 0.45 0.45 0.45 0.00 0.00 0.00 0.35 1974 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1975 0.00 0.00 0.00 0.00 0.00 1.02 1.02 1.02 0.15 1976 0.00 0.00 0.00 0.00 0.00 1.01 1.01 1.01 0.15 1977 0.00 0.00 0.28 0.28 0.28 0.52 0.52 0.52 0.30 1978 0.00 0.00 0.00 0.00 0.00 1.08 1.08 1.08 0.16 1979 0.00 0.00 0.00 0.00 0.00 0.49 0.49 0.49 0.07 1980 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1981 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1982 0.26 0.26 0.31 0.31 0.31 2.72 2.72 2.72 0.67 1983 0.00 0.00 0.00 0.00 0.00 0.00 0.00 2.61 0.08 1984 0.00 0.00 0.00 0.00 1.06 1.16 1.61 0.00 0.69 1985 0.00 0.00 0.00 0.00 0.00 0.00 1.63 3.77 0.19 1986 0.00 0.00 0.00 0.00 4.82 0.88 1.20 3.44 2.63 1987 0.00 0.00 0.00 0.00 0.00 3.73 0.95 2.95 0.40 1988 0.00 0.00 0.00 0.00 0.00 0.00 0.00 2.59 0.08 1989 0.00 0.00 0.00 0.80 1.07 0.79 1.82 4.71 0.94 1990 0.00 0.00 0.00 0.00 0.00 2.67 2.82 3.92 0.45 1991 0.00 0.00 0.76 0.00 0.00 1.06 0.00 9.89 0.47 1992 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.74 0.02 1993 0.00 0.00 0.00 0.00 0.00 0.81 0.00 0.75 0.08 1994 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.59 0.02 1995 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1.72 0.05 1996 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1997 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.47 0.01 1998 0.00 0.00 0.00 0.32 0.00 0.00 0.61 1.09 0.11 1999 0.00 0.00 0.00 0.00 0.34 0.47 0.00 2.27 0.27 2000 0.00 0.00 0.29 0.00 0.98 0.91 0.66 1.51 0.67

Source: Default and Recovery Rates of Corporate Bond Issuers: 2000 and Salomon Smith Barney.

JULY 2002

Glen McDermott Sevag Vartanian

The CDO Trustee Report Peeling Back the Layers



Introduction

Arbitrage cash flow CDOs are dynamic. The manager has the unfettered ability to trade credit-risk and credit-improved securities, as well as the right to trade 20% to 30% annually for any reason at all. This trading activity, coupled with the high percentage of ratings downgrades and defaults that the market has been experiencing over the last few years, has resulted in a pools of assets that can change significantly from month to month. From a risk management perspective, the information in a trustee report contains important early warning signs for investors. The information gleaned from the report should not be the beginning and end of a risk management program, but it should be an important and fundamental first step.

The typical CDO trustee report exceeds 100 pages, but there are usually only a small number of pages in the report that contain the most salient points. The goal of our brief report is to help new CDO investors negotiate their way through this dense document. Understanding and tracking the key metrics as reported in the trustee report is the first step in developing an effective risk management system; equally important, though, is an understanding of what a trustee report does not report. These omissions must be supplemented by conversations with the CDO managers and ultimately a robust analytical model. Only when armed with this information and analytics can an investor make an informed decision whether to buy, hold, or sell.

Overview of Arbitrage Cash Flow CDO Structures Collateralized debt obligations (CDOs) are formed when asset-backed structuring technology is applied to a pool of corporate credit exposures. CDO structures can be segmented into three categories: cash flow, market value, and credit derivative. Cash flow CDOs rely upon the cash flow generated from the pool of assets to service the issued debt. This report will attempt to explain how to decipher the typical cash flow CDO trustee report and also how to identify key indicators of performance.

A CDO is created when a special purpose vehicle (SPV) is established to acquire a pool of high yield corporate bonds, bank loans, asset-backed securities, or other debt obligations. In order to fund the acquisition of the debt obligations, the SPV issues rated and unrated liabilities. Since the majority of these liabilities are highly rated, the CDO can raise most of its capital cheaply in the investment-grade market and invest it most profitably in other markets, including those mentioned here.



Figure 111. Typical CDO Structure

COLLATERALMANAGER

TRUSTEE

ISSUER(SPV)

HEDGECOUNTER-

PARTY

CREDITENHANCER

SELLER

HIGH YIELDBOND

PORTFOLIO

INVESTORS

SUBORDINATEDNOTES

INCOME NOTES

SENIOR NOTES

BOND INTERESTAND PRINCIPAL

CDO NOTES

BOND PORTFOLIO

BONDPORTFOLIO

CDONOTE

PROCEEDS

CDO NOTE PROCEEDS

CDO NOTE PROCEEDS

Seller: Sells the bond portfolio to the issuer.Issuer: Issues CDO notes and uses note proceeds to buy bond portfolio.Investors: Purchase CDO Notes.Collateral Manager: Manages bond portfolio.Trustee: Fiduciary duty to protect investors’ security interest in bond portfolio.Hedge Counterparty: Provides interest swap to hedge fixed/floating rate mismatch.Credit Enhancer: Guarantees payment of principal and interest to note holders. Optional.


In a typical cash flow CDO, the rated liabilities are tranched into multiple classes, with the most senior class receiving a triple-A or double-A rating and the most subordinated class above the income note receiving a double-B or single-B rating. The ratings on the classes are a function of subordination and how cash flow and losses are allocated among them. Principal and interest cash flow are paid sequentially from the highest-rated class to the lowest, but if the cash flow is insufficient to meet senior costs, or certain asset maintenance tests are not met, most or all of the cash flow is paid to the most senior class.

There are two main phases during the life of an arbitrage cash flow CDO: the reinvestment period and the amortization period. During the reinvestment period (which typically lasts five years), only interest is paid to the CDO notes and the collateral manager reinvests any principal collections into new assets. The amortization period follows the reinvestment period, and during that time all principal collections are used to amortize the notes sequentially from the most senior to the most junior. There are two ways, however, that the reinvestment period can cease prematurely. All arbitrage cash flow CDOs contain two types of coverage tests: an asset coverage test (principal coverage test) and a liquidity coverage test (interest coverage test). If either test is violated, reinvestment of principal ceases and principal and interest collections are used to accelerate the redemption of the senior notes until the tests are brought back into compliance.

In most transactions, the principal and interest coverage tests are the only tests that result in the diversion of cash flow to the most senior notes outstanding. All other tests are collateral quality or “soft” tests that address concerns such as weighted average credit quality, obligor and sector concentrations, maturity limitations, and geographic



restrictions. If a “soft” test has been violated, the manager simply cannot make this level of noncompliance worse when he or she trades bonds in the underlying portfolio.

Coverage Tests These are the two most important tests in a CDO. A CDO is analogous to a small finance company, and like any company there is a formula that compares the value of its assets to its liabilities (principal coverage test) and a formula that compares its liquidity with its debt service (interest coverage test). The results of these tests are displayed on the cover page of the trustee report and the formulas can be found a few pages into the report. The typical principal coverage test is calculated as follows:

Calculation

Class A X/A

Class B X/(A+B)

The variable X represents the aggregate par amount of assets in the portfolio minus the aggregate par amount of defaulted assets plus the recovery value assigned to the defaulted assets plus the aggregate amount of principal proceeds held in the collection account. With respect to the denominator, the variable A (or B) represents the notional amount of the Class A (or Class B) notes. An overcollateralization ratio is computed for each class and for each class the denominator of the ratio is equal to the sum of the par amount of the class in question and all classes senior to the class in question. Violation of the principal coverage test of any class will result in the redirection of cash flow to pay down the most senior class then outstanding until the test is brought back into compliance.

When reviewing reported principal coverage tests each month, do not consider them in isolation — they must be analyzed in the context of the overall credit quality of the portfolio and the trading activity of the manager. For example, a deal may have a comfortable buffer between its actual principal coverage ratio and the trigger, but its portfolio may be of very poor credit quality as reflected in a high weighted average rating factor (see the “Weighted Average Debt Rating Test” section). A high weighted average rating factor is a leading indicator of future defaults and these future defaults will lower the principal coverage ratio, all other things equal. The buffer, therefore, may be illusory. Also, one must ask how a manager is building par. With respect to the manager’s trading activity, it can manipulate or “game” the principal coverage test by purchasing deeply discounted assets (these assets are included in the principal coverage ratio at par). Ultimately, many deeply discounted assets are trading that way for a reason: The market perceives heightened credit risk with respect to these names. Finally, even if a principal coverage test has been triggered, only the cash flow that is in the collection account on the calculation date (a trust accounting date that occurs just before each semi-annual distribution date) is eligible to delever the transaction.

Generally speaking, the margin by which an investment-grade deal passes the senior and subordinate principal coverage (OC) and interest coverage (IC) tests is less than the same margin for high yield bond deals. With respect to loan deals, the OC test margins are similar to those of high yield bond deals while the IC test margins are



much wider. The failure of any OC test should be a cause of concern.82 It is less of a concern, though, for loan deals, as the excess interest that loans generate, combined with higher loan recovery rates, allows the manager sufficient flexibility to recover. Of course, there is a point of no return, even for loan deals.

Collateral Quality Tests As we mentioned, when a collateral quality or “soft” test has been violated, the manager simply cannot make this level on noncompliance worse when he or she trades bonds in the underlying portfolio. The typical arbitrage cash flow CDO has some or all of the following collateral quality tests:

➤ Moody’s weighted average rating test;

➤ Weighted average life test;

➤ S&P weighted average recovery test;

➤ CCC concentration limits;

➤ Country/region concentration limits;

➤ Industry concentration limits;

➤ Single obligor concentration limits;

➤ Floating-rate asset limits;

➤ Maturity limits; and

➤ Limitations on certain types of securities (e.g., synthetics).

Although the remedy for violation of these tests is not as strong as deleveraging the portfolio, investors should nevertheless keep a close eye on these tests. They may foreshadow future problems in the portfolio (e.g., a rising CCC bucket) and they represent a starting point for investors to begin a dialogue with the collateral manager regarding its strategy for managing an ever-changing portfolio.83

Two of these collateral quality tests bear special mention. First, with respect to default rates by rating category, Figure 116 shows the cumulative annual default rates compiled by Moody’s from 1983–2001. Of all the factors noted, we find that it is most important to consider the percentage of CDS that are rated Caa1 and below. For a high yield CBO, once the percentage of assets rated Caa1 and below reaches 5%, investors should be aware and slightly concerned. Once it goes past 8%–10%, mezzanine holders should consider trading out with mild losses. As the CCC bucket grows past 10%, investors should be aware that losses can be substantial, as the default rate in this ratings category is so high. Of assets rated Caa1 or lower, more than half (57.44%) have defaulted over five years and about 80% have defaulted over ten years.84

82 When these deals are structured, the structuring agent always attempts to achieve the optimal amount of leverage in the capital structure and this usually results in less “buffer” in the OC ratio as compared to the IC ratio. As a result, if a deal is failing both its IC and OC test it generally is in bad shape. If this occurs, there is no excess interest to cure violations of the OC test.

83 In the standard Chase trustee report, these tests are found in the “Portfolio Percentage Limitations” section.

84 The typical cash flow CDO has a legal final maturity date of about 12 years.



Finally, with respect to single-issuer concentration limits, because CDO assets pools are lumpy to begin with, the more names in the pool, the less any single obligor default can hurt note holders. This is especially true with respect to the typical investment-grade corporate or ABS CDO, given their greater leverage relative to a high-yield CDO. In our opinion, obligor diversity is even more important than industry diversity, and we typically look for an investment-grade CDO to have at least one hundred different names with maximum position sizes between 1.0% to 1.25%. High yield CBOs can carry slightly larger positions. To determine the effect of obligor diversity on default volatility, we ran pricing simulations by varying the number of obligors in the underlying pool. The conclusion: As the number of bonds is reduced below 100 (assuming all are of equal size), default rate volatility increases dramatically (see Figure 112).85

Figure 112. Benefits of Obligor Diversity

0

1

2

3

4

5

6

7

8

9

10

10 100 1,000 10,000Number of Bonds in Colateral Portfolio

Cum

ulat

ive

15-Y

ear D

efau

lt Ra

te (%

) Mean

Std Dev

Mean cumulative default rates and standard deviations as a function of number of bonds in the collateral pool. Source: Salomon Smith Barney.

Weighted Average Debt Rating Test As Moody’s has commented “there are two broad credit considerations that determine the risk of a CBO or a CLO portfolio: collateral diversity by industry and by obligor, and the credit quality of each asset in the portfolio.” With respect to the latter, Moody’s uses the concept of the weighted average debt rating. Each security in the CDO portfolio has a rating (actual or implied) and each rating has a corresponding rating factor. The lower a security’s rating, the higher its corresponding rating factor. In order to calculate the weighted average debt rating for a pool of assets, take the par amount of each performing asset and multiply it by its respective rating factor. Then, sum the resulting amounts for all assets in the pool and divide this number by the sum of the par values of the performing assets.86

85 For a more in-depth discussion of this experiment please see Investment Grade CDOs, Salomon Smith Barney, McDermott, Benzschawel, and Khan, August 2001.

86 Modern CDOs exclude defaulted assets from the weighted average debt rating calculations, but some of the older deals (i.e., pre-2000) do not.



Figure 113. Moody’s Rating Factor Equivalents Rating of Debt Security Rating FactorAaa 1Aa1 10Aa2 20Aa3 40A1 70A2 120A3 180Baa1 260Baa2 360Baa3 610Ba1 940Ba2 1,350Ba3 1,780B1 2,220B2 2,720B3 3,490Caa 6,500Ca 10,000C 10,000

Sources: Rating Cash Flow Transactions Backed by Corporate Debt 1995 Update and Moody’s Investors Service, 1995.

The typical investment-grade average CDO is structured with a weighted average debt rating of Baa3, and this corresponds to a weighted average rating factor of 610. By contrast, many high yield CDOs have been structured with a weighted average debt rating of B2, which corresponds to a weighted average rating factor of 2720. What does a rating factor of 2720 imply about the default probability of the underlying portfolio? An investor can get a rough idea as to a portfolio’s ten-year cumulative gross default rate by moving the rating factor’s decimal over two spaces to the left. A rating factor of 2720, therefore, would imply a ten-year gross default rate of 27.20%.

As we discussed in the Coverage Tests section, when reviewing a CDO trustee report, investors must be careful not to analyze any performance metric in isolation. For example, a large month-to-month drop in the weighted average debt rating may appear to be a positive development at first glance. A reduction may signal an improvement in credit quality, but it may also reflect the sale of a poorly rated asset (which would almost certainly result in a loss of par). Since defaulted assets are typically excluded from the debt rating calculation, a security’s default may also cause this measure to drop.

Finally, the weighted average debt rating (by definition) gives investors no insight as to the dispersion of credits in the portfolio. Two portfolios may have the same weighted average debt rating, but they may have very different CCC baskets. One portfolio may have its credits tightly clustered around the portfolio’s mean debt rating and the other portfolio may be barbelled. To validate this point, we analyzed the weighted average debt ratings and triple C baskets in 255 high yield arbitrage CBOs and CLOs rated by Moody’s from 1996 to 2001.87 Although we observed a significant correlation between the weighted average debt ratings and the triple C baskets in a high yield CDOs in the 1998, 1999, and 2001 vintages (correlation coefficients of 0.59, 0.80, and 0.87, respectively), the correlation was not that

87 Collateralized Debt Obligations Performance Overview Compilation, Moody’s Investors Service, June 13, 2002.



pronounced in the other vintages. Namely, the correlation coefficient was only 0.05 for 2000 vintage and -0.01 for combined 1996 and 1997 vintages.

Diversity Score Diversity score is a Moody’s measure that reflects a portfolio’s diversity by industry and by obligor. The diversity score is one of the key components that Moody’s uses when assessing portfolio credit risk (binomial expansion technique or BET) and it is the way Moody’s takes industry-related default correlation into account.

Figure 114. Moody’s Diversity Score Number of Firms in Same Industry Diversity Score1 1.002 1.503 2.004 2.335 2.676 3.007 3.258 3.509 3.7510 4.00>10 Evaluated on a case-by-case basis

Source: Rating Cash Flow Transactions Backed by Corporate Debt 1995 Update, Moody’s Investors Service, 1995.

The binomial expansion technique uses the diversity score to construct a hypothetical pool of uncorrelated and homogeneous assets that are intended to mirror the behavior of the actual underlying portfolio. Figure 114 shows that the first firm in a sector earns the CDO one point towards the overall diversity score, but if there are two firms in the same sector, the contribution to the portfolio’s diversity score is 1.5, not 2.0.88 Default correlation within a sector counterbalances some of the benefits of portfolio diversity. Nevertheless, the higher the diversity score, the higher the obligor and/or sector diversity in the underlying portfolio and the lower its expected loss (see Figure 115).

Figure 115. Expected Loss Versus Diversity Score

0.0%

0.2%

0.4%

0.6%

0.8%

1.0%

1.2%

1.4%

1.6%

1.8%

5 10 15 20 25 30 35 40 45 50Diversity

Expe

cted

Los

s


88 The total diversity score for the transaction is calculated by summing the diversity scores calculated for each sector in the portfolio.



While the diversity score is a handy relative measure of portfolio obligor and sector diversity, it has its limitations. First, it is a static measure that is not terribly meaningful on its own. In order to calculate a portfolio’s expected loss, the diversity score must be analyzed along with the portfolio’s weighted average debt rating in the BET framework. On its own, one cannot conclude that a portfolio with a high diversity score is less risky than a portfolio with a lower score. Second, at least with respect to corporate CDOs, the diversity score assumes approximately a 30% correlation among names in an industry (irrespective of which industry) and no correlation across industries. In reality, this correlation factor may be too high or low depending on the actual industries in the underlying portfolio. The actual industries in the portfolio will also dictate how high (or low) the inter-industry default correlation will be.

Maintaining Par — Defaults, Purchases and Sales As a CDO investor, it is extremely important to keep track of any loss of par that occurs and, similarly, any gains in par. This analysis should go hand-in-hand with keeping track of trading activity. It is very rare that a CDO doesn’t give any telltale signs of impending doom. Although the use of constant default rate scenarios may be appropriate at the new issue stage of a deal, the vast disparity in default rates among rating categories becomes apparent as ratings fall lower and lower. Figure 116 shows the cumulative annual default rates compiled by Moody’s from 1983–2001.

Figure 116. Moody’s Cumulative Annual Default Rates, Based on 1983–2001 Sample Year 1 2 3 4 5 6 7 8 9 10Aaa 0.00% 0.00% 0.00% 0.07% 0.22% 0.31% 0.41% 0.53% 0.53% 0.53%Aa1 0.00 0.00 0.00 0.23 0.23 0.38 0.38 0.38 0.38 0.38Aa2 0.00 0.00 0.06 0.19 0.42 0.51 0.61 0.73 0.88 1.05Aa3 0.05 0.09 0.16 0.24 0.34 0.46 0.46 0.46 0.46 0.58A1 0.00 0.02 0.27 0.43 0.54 0.67 0.73 0.86 0.93 1.02A2 0.04 0.10 0.28 0.57 0.77 0.98 1.12 1.51 1.83 1.98A3 0.00 0.11 0.21 0.29 0.42 0.64 0.96 1.03 1.22 1.34Baa1 0.12 0.40 0.69 1.10 1.52 1.81 2.16 2.37 2.50 2.50Baa2 0.09 0.39 0.76 1.46 2.18 2.98 3.68 4.20 4.96 5.87Baa3 0.37 0.88 1.51 2.47 3.26 4.40 5.57 6.72 7.45 8.33Ba1 0.62 2.03 3.68 5.83 7.67 9.51 10.76 11.99 12.73 13.85Ba2 0.62 2.43 4.75 7.33 9.55 11.27 13.29 14.81 15.96 16.33Ba3 2.43 6.81 11.95 16.64 21.04 25.46 29.23 33.25 37.12 39.80B1 3.47 9.81 15.99 21.64 27.26 32.49 38.27 42.19 45.98 49.66B2 7.18 15.65 22.96 28.87 33.57 36.80 39.43 41.18 42.33 43.76B3 12.45 21.81 29.63 35.80 41.13 45.05 47.94 52.04 55.72 57.35Caa1 – C 21.61 34.23 44.04 52.18 57.44 62.52 66.37 71.17 75.61 80.49

Based on 1983–2001 Sample. Source: Moody's Corporate Default Study: 2001, Moody’s Investors Service, 2002.

Closing portfolios rarely have assets rated B3 or below. But as the table demonstrates, default rates in this area of the credit curve vastly exceed 2% CDR. Analysis of portfolios that have changed dramatically in credit quality from the closing date should incorporate some sort of more detailed default scenarios. Only in this manner can one determine what the expected loss of par will be.



The CDO manager is charged with preserving a portfolio’s aggregate principal par amount and maintaining stable credit quality. These two issues are, of course, related. If the aggregate principal par amount of a portfolio is stable, the transaction will generate sufficient cash flow to service the rated CDO notes and deliver solid returns to the CDO equity holders. There are two ways that an arbitrage cash flow CDO portfolio can lose par — through defaults and sales of assets at a discount to par (credit-risk sales).

With respect to defaults, a trustee report will generally show the defaulted assets that are currently held by the CDO, the default date for each asset, its market value and its assumed recovery value.89 For the purposes of the principal coverage test(s), each defaulted asset is carried at the lower of its current market price and its assumed recovery value as set forth in the indenture. Many CDO trustee reports, however, do not clearly and concisely state the defaults that the CDO has suffered over its life and the price at which the manager bought and sold each defaulted asset. The information is usually reported at some time during the life of the CDO, but it can be time-consuming and tedious to gather this information from the numerous monthly trustee reports. Despite these reporting limitations, we can gauge the overall effect of defaults and credit risk sales on a CDO portfolio by calculating its net annual default rate (NADR).90

Regarding sale activity, most trustee reports will show the reason for the sale (credit-improved, credit-risk, or discretionary) and the price at which each asset was sold. This information is valuable because it enables market participants to infer certain things about the effectiveness of the manager. Namely, by comparing the dollar price at which the manager executed each credit risk sale with the current market price of the sold position, an investor can see whether trading is being used to effectively mitigate losses.

With respect to purchase activity, investors should keep a close eye on each asset’s purchase price. A pattern of buying deeply discounted assets may indicate a more volatile trading strategy and may ultimately be inconsistent with the best interests of CDO note holders, especially rated note holders. At the very least, it should prompt a call to the manager.

Sometimes, a portfolio manager will attempt to support the principal coverage tests through risky trading strategies.91 Although it is impossible to have a credit opinion on every high yield trade that takes place, it is relatively simple to examine the prices at which bonds are being purchased and sold. One strategy that some managers execute is selling assets with a high dollar price and purchasing assets with a lower dollar price. Regardless of the price at which a security is purchased, the manager gets credit for the par amount of the collateral when the principal coverage test is calculated. So, if the amount of assets is $100 million and the manager sells $5

89 In the standard Chase report, this information can be found under “Portfolio Assets — Defaulted Securities.”

90 We do this by comparing the portfolio’s current performing par balance (performing assets plus principal cash) with its targeted principal balance at closing or the end of the ramp-up period (if there was a ramp-up period). The difference is equal to the loss of par that the portfolio has sustained over its life and it can be converted into an average annual default rate. This quick calculation can be used as a relative measure when comparing losses across CDOs. The NADR is not comparable to a single-year default rate for two reasons: (1) it gives credit to any realized recoveries and (2) it states the loss of par as a percentage of the original targeted balance. For these reasons, the NADR will be understated when compared to single-year gross default rates as reported by Moody’s.

91 See “CDO Spotlight: Par-Building Trades Merit Scrutiny,” Standard & Poor’s RatingsDirect, July 15, 2002.



million of a bond at $102, and purchases $10 million of a bond at $51, he or she has improved the OC test by 5%. Without having a credit opinion of the security purchased, it is probably safe to assume that any bond trading at that steep of a discount is more risky than the bond trading at a premium.

It is much more desirable to see trades where credit improved securities are sold along with those that have deteriorated, with the hope that the manager is actively trading to mitigate future losses and maintain par. A good strategy is to take small losses early instead of realizing larger losses later. This being said, one should understand that because of the narrower spreads in the investment-grade bond market, it is easier to recoup losses in investment-grade deals than it is in high yield deals. In addition, because of the high recovery rates in the high yield loan market, it is more difficult to lose par, and just as difficult to increase par. While opinions vary on whether active trading is preferable to passive investing, it is indisputable that a strong research team is crucial in determining the initial allocation of capital.

Finally, knowing what the manager’s experience has been in defaulted assets is also quite important. If the average amount of recovery for defaulted assets is 5%, what assumptions should one use for projected defaults? If the experience is 45%, should one feel more comfortable? These are qualitative questions that only the investor can answer. Past performance may not be indicative of future results, but the realized recovery rate is another data point that should be considered.

Subordination Defined There has been a fair amount of confusion in the market concerning the amount of subordination available to note holders in a cash flow arbitrage CDO. When reviewing a CDO trustee report, one of the first things that investors try to determine (frequently in the context of reviewing CDO positions in the secondary market) is how much subordination is beneath the class in question? Some of the answer can be found in the trustee report, but the remainder can only be ascertained by running a cash flow model.

CDO note holders enjoy two types of protection: principal subordination and excess spread. Measuring principal subordination is fairly straightforward. For any class, take the aggregate notional amount of the assets and subtract the notional amount of the class in question plus the notional amount of any classes senior to the class in question. Figure 117 provides a useful example.

At closing, the CDO in Figure 117 owned $500 million par amount in assets, and the amount of principal subordination available to the Class A-2 would be calculated as follows:

($500,000,000) – ($37,000,000 + $402,000,000) = $61,000,000

The principal amount of collateral subordinated to the Class A-2 notes is 12% of the aggregate collateral balance. If a Class A-2 investor, however, simply summed the notional amount of liabilities subordinate to the Class A-2 and divided it by the aggregate notional amount of the liabilities, this would understate the amount of principal protection available — ($27 million + $18 million)/$484 million = 9.3%. The second way to calculate principal protection understates the level of protection



and is incorrect because the assets were purchased at a discount. Clearly, this faulty method would overstate the amount of principal protection in the unlikely event that the assets were purchased at a premium. For example, with respect to the most subordinated tranche above the equity, if the amount of assets in the portfolio plus principal cash is equal to the aggregate amount of the rated notes, this implies that the subordinated tranche has become the first loss piece, and the equity notes have lost most, if not all, of their value.92

Figure 117. Sample CDO Capital Structure Cum. Amt. of Pari Collateral in Excess of Debt

Tranche Principal Amount Passu or Senior Debt Collateral RatioClass A-1 $402,000,000 $402,000,000 $98,000,000 124.4%Class A-2 37,000,000 439,000,000 61,000,000 113.9Class B 27,000,000 466,000,000 34,000,000 107.3Equity 18,000,000 484,000,000 16,000,000 —


Excess spread, the second type of protection to CDO note holders, is harder to quantify. Excess spread benefits note holders to the extent that it is used to amortize the liabilities, rather than being paid out to the CDO equity holders. This redirection of cash flow occurs when an interest or principal coverage test is tripped. A number of key variables will determine whether these tests are in compliance:

➤ The size of the buffer above the minimum interest or principal coverage tests;

➤ Trading gains and losses;

➤ The level and timing of defaults and recoveries; and

➤ The yield on the portfolio.

Because a number of variables will affect the value of excess spread to note holders, the best way to quantify it is to run numerous cash flow scenarios. For example, in a front-loaded default scenario with lagged recoveries, the amount of excess spread available to rated note holders would increase because it would be one way to hit the triggers and redirect cash flow from the CDO equity to the most senior notes then outstanding. Taking the present value of all redirected cash flows would be one way to quantify the benefit.

Conclusion We have discussed some of the fundamental features common to all arbitrage cash flow CDO trustee reports. The information supplied in this report forms the foundation upon which an effective risk management system can be built. This is the beginning of such a program, not the end. Tracking CDO trustee report metrics is part of a larger puzzle: robust cash flow models and conversations with the manager are additional pieces.

92 When analyzing CDOs in the secondary, many investors assume all triple-C exposures default over a two-year period with very low realized recovery rates.



Credit Derivatives

APRIL 27, 2004

David Li Ratul Roy Jure Skarabot

.

A Primer on Single-Tranche CDOs



Su

mm

ary

Single-Tranche CDOs

The Single-Tranche CDO (STCDO) product represents the newest generation of tranching technology. It gives investors exposure to a customized slice (tranche) of the credit risk of a selected portfolio of reference credits. Although it is usually structured in an unfunded form as a portfolio default swap, it can also be designed as a funded obligation through a credit-linked note. The advent of STCDOs has improved the liquidity, transparency and flexibility of tranched products.

Decision Steps for Investors

Investors in STCDOs have the flexibility to choose the credit portfolio that they want exposure to, or they may reference the STCDO to the liquid iBoxx index. Investors’ desired rating and return target determines their choice of subordination level and tranche size. In addition, investors can substitute underlying credits in the portfolio at the prevailing market prices throughout the life of a single-tranche CDO.

Key Issues in Modeling and Valuation

The two key elements in the modeling and valuation of structured credit portfolio products such as single-tranche CDOs are single-name default probabilities for the entities in the reference portfolio and the default correlation among the credits. The standard model that has been adopted by the market is the Gaussian Copula model, which, when combined with Monte Carlo simulation, generates correlated default time scenarios and a loss distribution for a STCDO tranche.

Risk Measures, Hedging, and Substitution of Credits

Investors primarily focus on five risk measures: credit spread sensitivity, default sensitivity, convexity, default correlation, and time-decay sensitivity. Junior tranches have greater credit spread sensitivity and default sensitivity than senior tranches, while junior and senior tranches have opposite convexity properties. An increase in default correlation is positive for junior tranches and negative for senior tranches. Time decay is caused by mismatches between the changing rate of default-related losses and the (constant) premium stream.

The ability to substitute credits allows investors to stave off imminent credit events, or to tweak the yield and risk characteristics during the life of the STCDO. While the exact cost of substitution is specified by the copula model, a simple approximation lets investors ballpark the fair value of substitution.

Single-Tranche CDO Investment Strategies

Single-tranche CDOs are popular among investors because they can be used to execute a variety of customized investment objectives and strategies. Single-tranche CDO strategies can be classified into several categories: leverage, market view, correlation, relative value, and micro/macro hedging. We describe each strategy, explain the market motivation for each, and indicate the typical investor who is likely to employ a given strategy.

.



Single-Tranche CDOs

Overview A single-tranche CDO (STCDO) is a powerful and flexible vehicle for investors seeking credit exposure consistent with their risk-return preference. The single-tranche CDO gives investors exposure to one slice (tranche) of credit risk of a selected portfolio of pre-selected reference credits. Investors may choose the credit portfolio that they want exposure to as well as the specific part or tranche of the capital structure. Investors can also choose a customized subordination level and tranche size. Investors with a higher rating requirement and a lower yield target can choose a greater level of subordination and/or a larger tranche size. The flexibility of the product is enhanced by investors’ ability to substitute underlying credits in the portfolio at the prevailing market prices during the life of the investment (see Figure 118 for the basic structure of single-tranche CDO).

Figure 118. Basic Format of a Single-Tranche CDO

Universe of Credits

ReferencePortfolio

TrancheSize

SeniorTranche

SelectedTranche

SubordinatedTranche

Investor

AttachmentPoint

Source: Citigroup.

Advantages of Single-Tranche CDOs

Single-tranche CDOs have several advantages versus alternative investment products. First, the leveraged tranche structure often provides higher yields than a similarly rated corporate investment. Second, investors have the ability to customize trades in terms of portfolio selection, the choice of subordination level, and tranche size. STCDOs also provide efficient market access to credits that might be expensive and/or difficult to acquire in the underlying cash markets. STCDOs give investors the ability to easily and efficiently go short the selected portfolio, sectors, and specific tranches. Investors can dynamically manage their positions through substitution of credits in the reference portfolio through the life of the single-tranche CDO. Finally, using single-tranche CDOs, investors can more efficiently hedge their portfolios, separating the fundamental (default) risk from the market-wide risk factors.

Single-tranche CDOs allow investors to take

an exposure to a specific tranche of

portfolio credit risk.

Single-tranche CDOs provide multiple

customized solutions for investors.



Key Features of Single-Tranche CDO Transaction

The key value drivers of single-tranche CDOs are the credit quality of the underlying portfolio, the subordination level and tranche size, and the correlation structure among the credits in the portfolio. Other features, such as substitution rights, also play an important role. Single-tranche CDO investors range from buy-and-hold investors who desire limited substitution rights and tend to hold the investment to maturity to correlation traders who use single-tranche CDOs to pursue various long and short strategies. However, traders are not the only ones who use STCDOs as defensive vehicles; for example, even traditional long-only investors can and do buy protection on single-tranche CDOs referenced to customized credit portfolios or the liquid iBoxx index to hedge fundamental (default) risk in their portfolios or to protect against market-wide spread selloffs.

Dealers take on risk when selling single-tranche CDO transactions to investors because they are placing only a portion of the capital structure. Dealers mitigate this risk by hedging with single name default swaps, default indices (e.g., iBoxx) or with other portfolio transactions. For example, “delta hedging,” offers dealers a way to manage their mark-to-market risk from underlying credit default swap spread movement. There are other risks that dealers need to manage as well, such as the risk of individual credits defaulting, correlation risk, shortening of maturity and convexity effects.

Description of the Product and Basic Structure A single-tranche CDO is a financial contract between two parties that references a portfolio of credits. The parties are called protection buyer and protection seller. Once a reference portfolio is specified, a single tranche is commonly referred to as an “x excess y” tranche, where x is the tranche size and y the subordination level (attachment point). Equivalently, the tranche covers losses between y and x + y part of the total portfolio losses. Protection sellers can invest in a STCDO either in a funded note or unfunded portfolio default swap format. For ease of exposition, for the remainder of this report we will generally use the term investor to denote the seller of protection, and we will explicitly point out the instances where the investor is the buyer of protection.

Key drivers for single-tranche CDOs are credit

quality of individual entities in the portfolio,

subordination level, tranche size, and the correlation structure.

A single-tranche CDO is a contract between protection buyer and

protection seller.



Figure 119. Mechanics of Single-Tranche CDO Transaction

Credit Risk Transfer

Premium Protection Buyer

Reference Portfolio $1 Billion

Protection covers lossesbetween 7%–10%

Payment if the loss exceedsthe subordination level, capped by the

tranche size

Protection Seller

7%-10%Tranche

Tranche size $30 Million

Source: Citigroup.

Single-Tranche CDO — Unfunded Form

If structured as a portfolio default swap, the investor receives a periodic premium usually expressed as a fixed percentage in basis points of the outstanding notional amount of the tranche. In return the investor provides protection for any loss more than subordination level, but the loss payment made by investors is capped by the tranche size, i.e., the maximum loss for an investor is the tranche size.93 Therefore, the cash flow exchanged between the two counterparties is default swap premium from protection buyer to the protection seller, and the loss payment, if any, from the protection seller to the protection buyer.

Single-Tranche CDO — Funded Form

Although the unfunded form is more typical, the same investment position can also be designed as a funded obligation through a credit-linked note. If structured in a funded note form, the notional amount the investor pays on closing is usually invested in high-quality, liquid assets such triple-A rated asset-backed securities. The note pays fixed rate or LIBOR plus a premium on the outstanding notional. At maturity, the investor is paid back the notional, unless the losses exceed the subordination level. If that occurs, the notional is reduced, and a portion of the collateral is liquidated and paid to the protection buyer.

Variations of the Standard Single-Tranche CDO Structure

There are several variations to the standard single-tranche CDO transaction. First, variations can be created on how the premium is paid. For instance, the protection seller may prefer to receive the present value of the premium upfront or part of the premium upfront (this is referred to as a “points upfront” payment and is usually used in the equity tranches). Other protection sellers may prefer premium payments to be increasing or decreasing over time to mitigate or reverse time decay effects (see

93 In a standard STCDO structure, it is easy to estimate the number of defaults in the underlying portfolio that will cause loss of payment by protection seller. As an example, assume that the tranche has 3% subordination, the reference portfolio has 100 names with $10 million size each, and the recovery rate is 40% for each credit. Five defaults can happen without affecting the tranche (five defaults will lead to $30 million loss on the portfolio (5 *$10 million*(100%–40%)), and the losses associated with the sixth default will be covered by the protection seller.

Single-tranche CDOs are commonly structured as

unfunded portfolio swap . . .

. . . and the same single-tranche CDO can also be structured as a

funded obligation.



the section “Single-Tranche CDO Risk Measures and Hedging” below). So far, most of the transactions have been based on a horizontal slice of the portfolio as represented in Figure 118, but there is no practical difficulty for dealers to offer a “vertical” slice of a tranche, such as 50% of the “3 excess 7” tranche, where the premium is paid based on half of the underlying notional amount, but the loss is based on the full amount of the tranche. Such tranche specification offers additional leverage and could be particularly interesting for investors selling protection on a pre-existing portfolio of credit obligations, where the original owner keeps the part of the “vertical” exposure in order to resolve potential adverse selection conflicts.

Second, the protection seller may want to have different parts of his investment exposed to different risk profiles. For example, the investor might want the final payoff of his investment principal linked to a single highly rated credit and the coupon payments to the junior STCDO tranche. Such variations, made possible by STCDO technology are similar in principle to a principal-protected structure, which allows risk-averse investors to enjoy the upside in a risky investment.

Third, the flexibility of single tranche CDOs allows investors to better manage their investment or exposures in conjunction with other business objectives. For example, a life insurance company might prefer to receive a decreasing premium payment scheme to increase its near-term earnings profile and better match its long-term asset-liability mix.

Ramp-up Period for Single-Tranche CDOs

The single-tranche CDO structure resolves another structural conflict inherent in traditional CDO investments. Traditional cash CDOs or balance-sheet synthetic CDOs are basically supply-driven products because brokers and dealers drive the reference portfolio selection and the ramp-up process. Unless they wish to carry a substantial amount of risk, the originators of a synthetic CDO are forced to sell the entire capital structure to investors. This process can be time-consuming and costly, often implicitly resulting in a conflict of interest between senior and junior investors. With only two parties to a single-tranche CDO, the buyer of a tranche (seller of protection) takes on the tranche credit risk and the protection buyer hedges mark-to-market risk in the liquid CDS market. This dramatically reduces the transaction period. Supply and demand for different parts of the capital structure are resolved through differences in the levels of implied correlation.

Main Decision Steps for Investors In a typical single-tranche CDO transaction, there are three main decision steps for potential investors:

1 Select a portfolio of credits to that they want exposure to.

2 Choose a subordination level (attachment point) and a tranche size corresponding to their risk/return preference or yield target.

3 Dynamically manage their position and substitute credits in the collateral portfolio throughout the life of a single-tranche CDO.

A single-tranche CDO can be structured in a fast and efficient way.



Selection of single-tranche CDO reference portfolio

The first step in structuring a single-tranche CDO transaction is the investors’ selection of the credits in the underlying reference portfolio according to their preferences. For example, investors can choose a portfolio of credits different from their current positions, and by selling the protection on those names they achieve further diversification of their overall credit exposure. They can also sell protection on the subset of names in their current portfolio in case they want to overweight certain credit or sectors. Alternatively, investors can designate the whole or a part of their portfolios as the reference pool, and by buying protection they hedge themselves against spread widening or individual defaults.

As a common alternative, investors choose iBoxx default index, which is based on a diversified set of liquid names in the credit default market as a reference portfolio. The main reason for that choice is the liquidity and diversification of the index. The standard tranches on the iBoxx index also trade as liquid instruments in the broker/dealer market.

Selection of the Subordination Level and Tranche Size

Once a portfolio is selected, investors must choose a subordination level and tranche size. The tranche size and subordination level determine the degree of leverage and the required protection premium. Investors that are primarily concerned about the rating and want to have the tranche rated by rating agencies such as Moody’s and S&P, could choose the tranche size and subordination level so as to minimize the premium paid for the selected rating. Other investors could choose a subordination level that would provide a total spread equal to some desired return target. This “tranching” of credit portfolio risk can provide any desired risk/return profile. The idea is similar to one used by insurers for many years; the subordination level is equivalent to the deductible in the insurance contract and the tranche size is analogous to the concept of insurance risk ceiling (“maximum amount of coverage”). Also, by choosing the position of a tranche on a capital structure of STCDO, investors can separate their views on default risk from their views on market risk. We review such investment strategies in a later section. These trades are analyzed in more detail in our recent publication, Bull and Bear in a Boxx.94

Substitution of Credits in the Reference Portfolio

The third distinctive feature of single-tranche CDOs is investors’ ability to dynamically manage their investment gain and loss by substituting credits in the portfolio. Because the credit risk of individual credits in the portfolio changes after the initial transaction, investors might wish to substitute certain names. They could, for example, eliminate issuers that they perceive as potential credit blowups and substitute them with less risky names. Following each change, the dealer will adjust the premium for the tranche, change the level of subordination, or settle through a upfront payment to offset the impact on the substitution on the mark-to-market of the tranche. Substitution rights are especially important for market participants who intend to use single-tranche synthetic CDO structures as an efficient hedging and portfolio-rebalancing tool to manage their existing credit portfolio. In case the portfolio has experienced losses and

94 Bull and Bear in a Boxx — Using Tranched Products to Express Credit View, Arvind Rajan, Graham Murphy, and Jure Skarabot, Citigroup, February 19, 2004.

Investors can select customized reference

portfolios or use default index (e.g., iBoxx) to sell

or buy default protection.

Subordination level and tranche size determine

the leverage and protection premium.

In a single-tranche CDO transaction, investor have

the ability to substitute the reference credits during

the life of investment.



the subordination below the invested tranches has been decreased, investors can improve the rating of their tranche by increasing the subordination level, even if they do not alter the composition of the reference portfolio.

Role of Outside Managers

Although substitution rights are inherently in the hands of the investors, some market participants prefer the delegate these tasks to an outside investors. Choosing a manager for a single-tranche CDO involves many of the same considerations that apply when picking a manager for a full capital structure synthetic CDO. In rating tranches managed by the third-party managers, the rating agencies pay special attention to the set of investment guidelines governing the managers’ trading behavior. Investors should be careful to ensure that the guidelines and rules are set up to avoid a conflict of interest with the fiduciary responsibilities of the manager to the investor.

Key Issues in Modeling and Valuation The cash flows of STCDOs depend only on the actual occurrence of defaults in the underlying credit portfolio. These cash flows are the overriding consideration for investors who are not subject to mark-to-market requirements and are expecting to hold the investment to maturity. However, investors who are subject to mark-to-market requirements, are also affected by the day-to-day spread movements of individual credit default swaps as well as the market perception of default correlation among the credits within the portfolio. Therefore, these two categories of investors would emphasize different risks when investing in STCDOs.

Any STCDO pricing model has the following two essential parts:

1 A source of information that provides single-name default probability for the credits in the reference portfolio.

2 A mechanism for the introduction of default correlation among credits.

Single-Name Information

The most commonly used sources of information for single-name default analysis are:

1 Historical information based on rating agency default experience studies.

2 Market and rating-driven spreads for market-traded instruments such as credit default swaps.

3 Debt-equity based measures of default probability — for example, Moody’s/KMV EDF model of PD (probability of default) from Citigroup’s internal debt-equity model95.

The choice of information source should be driven by investment objectives.

Historical information. This is used for performance and risk measurement by net risk takers interested in long-term credit exposure. However, these investors will tend to require additional premium over and above the compensation based on expected loss derived from historical data. This additional premium, much like “risk 95 See Jorge Sobehart and Sean Keenan, Hybrid Probability of Default Models — A Practical Approach to Modeling Default Risk, Citigroup, October 2003.

Two key issue elements for modeling and

valuation are single-name default

probabilities and default correlation structure.

Single-name default probabilities can be derived

from various sources.



loading” within the insurance industry, is based on a multiple of unexpected loss. This risk premium is associated with the investor’s risk preference and appetite for unexpected losses.

Spreads. Some investors use a combination of rating and spread such as a matrix of idealized credit spreads classified by rating and industry. Others may use historical default data, but rather than using the current rating of the credit, they may choose to work with the rating implied from current market spreads.

Broker-dealers tend to use market information because they need to manage the composite trading book (portfolio transactions and single-name default swaps) in a consistent way.

Debt-equity information. Debt-equity based information of default probability incorporates forward-looking credit measures of default expectation for a specific firm over a given time horizon. The default expectation is based on the Merton-type structural model of default risk. This framework is driven by two main inputs: the asset volatility and the leverage of the firm. The asset volatility of the firm is determined from market observables, such as equity price and equity volatility, and leverage information based on the firm’s financial information. Commonly used debt-equity model measures are EDFs developed by Moody’s/KMV.

Default Correlation

The information on single-name default risk alone is insufficient to characterize the default risk of a portfolio tranche. This is because of default correlation. Put simply, default correlation measures the degree to which the default of one asset makes the default of another asset more or less likely. The concept of correlation is conceptually related to the more widely understood concept of diversification in credit portfolios.

Intuitively, one can think of default risk as being driven by a set of factors shared to different degrees by the individual credits. These factors tend to tie all credits into a common set of economic risks, but to different degrees. A particular factor may be associated with the general economy, or a region, or a specific sector, or, more idiosyncratically, related to the specific credit itself. Due to the shared influence of certain factors, it is generally believed that default correlation is positive between companies, even if they operate in different sectors. The greater the influence of general factors on a particular company, the greater the correlation of the default risk of that company with the general market, and with other companies in the market.

Within the same sector we would expect companies to have an even higher default correlation because they have more factors in common. As an example, consider the recent overcapacity of the telecommunication industry, a sector-specific issue, which resulted in the default of numerous telecommunication and telephone companies.

Quantifying default correlation is a challenge because of scarce data, even though the standard analytical framework is well understood. This framework, which is based on copula analysis, requires the prediction of joint (that is, correlated) survival times for all credits within a reference portfolio. Individual default probabilities only provide survival time for single credits. A copula function provides a method of obtaining a joint or multivariate distribution from a set of univariate distributions.

Default correlation plays an important role in

modeling and valuation of structured credit portfolio products.



Model Implementation

Once a joint distribution of default times is obtained, a Monte Carlo simulation algorithm can be used to generate a large number of correlated default time scenarios. For each simulation scenario, we know exactly which credit defaults within the maturity of the transaction and exactly when that default occurs. By aggregating the results of all the scenarios, we estimate the loss distribution of the whole portfolio. The loss distribution of any particular tranche of a certain size with a subordination level can then be inferred from the portfolio’s total loss distribution. The break-even spread for a tranche is defined as the premium that equates the present value of expected losses with the present value of premium payments for the tranche. This is the most common way of implementing the copula function approach to credit portfolio modeling. Alternative approaches that use Fast Fourier Transform, recursive calculation, or conditional analytical approximation can substantially reduce the computational time and noise inherent in the Monte Carlo simulation.

Role of Correlation

The expected loss of a portfolio of credits is the sum of the expected losses of each credit. This sum is not affected by portfolio diversity or default correlation.

However, while leaving its mean unchanged, default correlation does affect the shape of the portfolio loss distribution, particularly its variance and its tail. As tranches are exposed to a slice of the portfolio loss, they are particularly sensitive to properties of the corresponding range of the loss distribution. Thus, although the portfolio expected loss is indifferent, their expected loss is sensitive to correlation via its effect on the shape of the loss distribution.

Correlation estimates can be obtained from historical default analysis or they can be viewed as a parameter implied by quoted tranche prices. Intuitively, increasing correlation has the effect of making extreme events — very few defaults or very many defaults — more likely, thus increasing the loss variance and the loss tail.

The standard model that has been adopted by the market is the Gaussian Copula model.96 In this version of the model, correlation is incorporated via a parameter that is approximately the correlation of issuers’ asset returns.

Single-Tranche CDO Risk Measures and Hedging Valuation models for single-tranche CDOs are important for more than just the pricing. They establish a platform to analyze the risks associated with the specific tranche investments. Based on the estimation of tranche risk sensitivities, buy-and-hold investors can estimate their mark-to-market exposure, and correlation traders can structure their hedging strategies. Note that different STCDO tranches on the capital structure can have a different direction and magnitude of sensitivity to each risk factor.

96 David Li: “On Default Correlation: A Copula Function Approach,” Journal of Fixed Income, March 2000, pp. 43–53.

The Gaussian Copula model is the standard

market approach to incorporate default

correlation in structured credit

portfolio modeling.

Correlation estimates can be derived from

historical analysis or can be used as market-

implied factor for the quoted tranche prices.

Risk measures (Greeks) estimate sensitivities of single-tranche CDOs to risk factors. Junior and

senior tranches on capital structure can have

different direction and magnitude of sensitivity.



The buyer or seller of a single-tranche CDO should consider the following risks measures (Greeks) 97:

1 Credit spread sensitivity (delta).

2 Credit spread convexity (gamma).

3 Correlation sensitivity (rho).

4 Default sensitivity (omega).

5 Time-decay sensitivity (theta).

Credit Spread Sensitivity — Delta

When the underlying credit spreads widen, the total portfolio expected loss will increase and correspondingly the expected loss of all tranches. A long position in tranches will thus see its mark-to-market decline and its breakeven spread rise as underlying spreads increase. There are two forms of spread sensitivity, individual (or micro) spread sensitivity delta, and sensitivity to a broad move in the portfolio spread, which is called the Credit 01.

We define the (individual) credit spread sensitivity delta as the ratio of the mark-to-market value change (∆ MTM) of a tranche Tj to the mark-to-market value change of a single name default swap when the spread of one individual credit CDSi moves by a small amount, typically one basis point.

∆MTM(CDSi)∆MTM(Tj)Tj

i ∆ =

Generally speaking, delta increases as we move down the capital structure, with the greatest delta (in absolute terms) in the equity tranche. In addition, spread delta for the equity tranche is higher for the individual credits with higher spread and spread delta for senior tranches is lower for the individual credits with higher spreads (see Figure 120). The reason is that when an individual credit spread is higher, the sensitivity of the senior tranche to this credit is lower as the specific credit is more likely to default and hit the equity tranche. For the same reason, spread delta for the equity tranche is higher for the credits with higher spread.

97 Default sensitivity has not yet been associated with a Greek letter. We use letter omega for default sensitivity.

Delta measures the change in tranche mark-

to-market value as the spread of one underlying

credit moves.



Figure 120. Individual Credit Spread Deltas as a Function of Individual Credit Spread Levels

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

15 65 115 165 215 265 315Spread (bp)

Spre

ad D

elta

(%)

15%–30% 10%–15% 7%–10% 3%–7% 0%–3%

Source: Citigroup.

In a similar fashion, we define the Credit 01 as the mark-to-market change in dollar value of a tranche if all the names in the portfolio widen by 1bp in a parallel move. Credit 01 is an aggregate spread sensitivity measure, and is thus a suitable measure for estimating the hedge ratio when, for example, delta-hedging an iBoxx tranche with the underlying iBoxx index. Unlike an individual spread sensitivity, Credit 01 of senior tranches increases as all spreads widen in a parallel way and Credit 01 of the equity tranche decreases if all spreads widen in a parallel way (see Figure 121). With a parallel spread shift, the senior tranches become more risky and therefore more sensitive to spread widening, and the equity tranche becomes less sensitive to additional spread widening.

Figure 121. Credit 01 Versus Parallel Spread Changes

0

10

20

30

40

50

60

70

80

-10bp -5bp unch +5bp +10bp +15bp +20bp +25bp +30bpSpread change

Cred

it 01

($M

)

15%–30% 10%–15%7%–10% 3%–7%0%–3%

Source: Citigroup.

Tranche Credit 01 for parallel spread moves

estimates the change in tranche value if all the names in the portfolio

widen by 1bp.



Credit Spread Convexity — Gamma

Just as with spread sensitivity, spread convexity comes in two flavors, macro- and micro (idiosyncratic or credit-specific).

Macro Spread Convexity

We define macro spread convexity as the additional mark-to-market change on a tranche over that obtained by multiplying the Credit 01 by the parallel spread move for the underlying single-name credit default swaps. In general the delta to the portfolio spread is greatest in the equity tranche. However, if spreads widen on average, the expected loss on the equity tranche draws nearer to the maximum loss it can sustain — tranche notional. For this reason, a long credit position in the equity tranche will have positive convexity (like a long bond position), and a long credit position in more senior tranches will have negative convexity. The best way to understand this is through an analogy with how simple options work. The tranche loss profile with respect to total portfolio loss resembles the payoff profile at expiration of a call spread with respect to the price of the underlying. Buying protection on a senior tranche is thus equivalent to owning an out-of-the-money call spread on default-driven losses in the underlying portfolio. The first few defaults do not affect the payout of tranche. As spreads go up, defaults become more likely, so the out-of-the-money option comes closer to becoming an at-the-money option, and the tranche sensitivity to spreads increases. Conversely, protection bought on a sufficiently junior tranche is either at-the-money or in the money. Thus an increase in spread reduces its sensitivity, as it goes deeper in the money.

Idiosyncratic Spread Convexity

Perhaps against our intuition, the convexity of a senior tranche to the spread of an individual name in the portfolio is generally of the opposite sign to the convexity with respect to the average spread. But this makes sense if you consider how the sensitivity of the portfolio spread to the changes in the individual credit spread is distributed amongst tranches. We observe that wide credits have a greater impact on junior tranches, and tight credits have a relatively greater impact on senior tranches. As the spread of an individual credit widens, junior tranches receive a greater portion of the credit’s delta and senior tranches receive less. Thus a long credit position in a senior tranche will have positive convexity to an individual credit spread.

Default Sensitivity — Omega

Following a credit event in the portfolio, the seller of protection on the equity tranche will compensate the buyer of protection for the loss suffered by the credit, and the notional amount of the tranche will be decreased by the one minus the recovery rate multiplied by the notional amount of the defaulted credit in the portfolio. The buyers of senior tranches will lose a portion of the subordination, and the tranches will become more risky. That will affect their mark-to-market positions.

We define Default 01 as the mark-to-market change in the value of a tranche if one of the credits defaults (the defaulted credit is chosen to have the largest impact on the tranche). We observe that the equity tranche is allocated most of the Default 01 risk in the portfolio, but its percentage share decreases as the spreads widen (see Figure 122).

A long position in an equity tranche has

positive spread convexity, and a long

position in senior tranches has negative

spread convexity.

Equity tranches take the largest part of the

default risk.



Figure 122. Distribution of Default 01 Risk Measure Unchanged Spreads Spreads Wider 25bp

18%

0%1%

76%

5%

15%–30%

10%–15%

7%–10%

3%–7%

0%–3%

22%

8%

3%

67%

0%

15%–30%

10%–15%

7%–10%

3%–7%

0%–3%

Source: Citigroup.

Correlation sensitivity (rho)

We understand that increased correlation makes more extreme events more likely. The increased likelihood of both “good” and “bad” extremes pushes expected loss out of junior tranches and into senior tranches, with the total expected loss unaffected. So, long equity tranche and short senior tranche credit positions have positive correlation sensitivity, and short equity and long senior tranche credit positions have negative correlation sensitivity. Also, with senior tranches gaining risk and junior tranches shedding it, there will also be a region in the capital structure that is relatively insensitive to a change in correlation. In a tight spread environment, this is generally the junior mezzanine tranche (see Figure 123).

Figure 123. Mark-to-Market Value Change Versus the Correlation Change

(1,500)

(1,000)

(500)

0

500

1,000

1,500

2,000

-15% -10% -5% unch +5% +10% +15%Correlation change

MTM

($M

)

15%–30% 10%–15%7%–10% 3%–7%0%–3%

Source: Citigroup.

Investors in senior tranches are

short correlation, and investors in

equity tranches are long correlations. Junior mezzanine tranche is usually

correlation insensitive.



Time-Decay Sensitivity — Theta Theta measures the impact on the value of tranches as the remaining time to maturity decreases. When we price a tranche, the breakeven spread is the spread, which would make the value of tranche loss equal to the total premium on expectation at inception. But as time elapses this equality will no longer hold. The reason is that the premium received for each period does not necessarily exactly offset the loss for the tranche in that period. The following graph shows the periodic expected loss and the expected premium over each period for the next five years for the iBoxx 3%–7% tranche (see Figure 124).

Figure 124. Periodic Expected Premium and Loss for Tranche (3%–7%)

0

100,000

200,000

300,000

400,000

500,000

600,000

0 5 10 15 20 25

Expected Loss

Expected Premium

Source: Citigroup.

This graph shows that at the beginning the protection buyers pay more than enough to cover the expected loss over each period for the first nine periods. However, that relationship reverses after the first nine periods. So the theta for this tranche from protection provider’s perspective is initially negative. This means that a long credit position in this tranche will initially suffer negative mark-to-market, like a premium bond. Similar graphs for other two tranches, 0–3% and 10%–15% are shown on the following figures (see Figures 125 and 126).


0

20,000

40,000

60,000

80,000

100,000

120,000

0 5 10 15 20 25Period

Amou

nt ($

)

Expected Loss

Expected Premium

Source: Citigroup.

Theta measures the impact of time-decay on the value of the tranche. The value of the tranche

changes as the time passes, because period

by period, premium received is not exactly offsetting the expected

tranche loss.




0

200,000

400,000

600,000

800,000

1,000,000

1,200,000

1,400,000

1,600,000

0 5 10 15 20 25Period

Amou

nt ($

)

Expected Loss

Expected Premium

Source: Citigroup.

For all senior tranches the periodic premium is flat, which reflects small incremental loss over each period while the expected periodic loss increases steadily. The equity tranche is the only tranche whose expected loss is larger than the expected premium at the origination of the transaction. In the above example (see Figure 126) the equity tranche premium is paid as the running spread. In most transacted cases, the market follows the “points upfront” convention, similar to the trading convention for the single-name CDS on the distressed names. In this case, the protection on the equity tranche is paid as percentage of the notional upfront and the fixed running premium (e.g., 500bp annual). Points upfront is defined as the difference between the present value of the expected loss on a tranche and the present value of expected fixed running premium. Therefore, as time passes, the value of points up front decreases, just as a bond priced initially at a discount pulls to par.

Substitution of Credits

Advantage of Substitution

When investing in static CDOs, the investor’s knowledge on the underlying credit pool is only used at the time that the investor signs the contract. After that, the investor can only hope for the best performance of his investment with no option to dynamically manage credit risk by replacing credits of the underlying portfolio. In contrast, with the managed single tranche, the investor is allowed to make a limited number of credit replacements per year. For example, an investor might be allowed to substitute ten names per year, with a maximum change of 50% of the portfolio composition during the maturity period of five years for a single tranche transaction with a reference credit pool of 100 names. This flexibility gives the investor freedom to manage the investment in a dynamic way. For instance, managed single tranches avoided a portion of the losses associated with the recent default of Parmalat because certain investors made the substitution of Parmalat when Parmalat was traded with an expected recovery rate of 20 or 30%, while the ultimate recovery turned out to be far lower.

If paid as running spread, equity tranche

premium is initially smaller than the

expected loss. In the market, equity tranche

default protection is usually paid as points upfront plus the fixed

running premium.

Investors in single-tranche CDOs have the

flexibility to replace credits in the underlying

portfolio during the life of the tranche.



Substitution Calculation

Since the dealer needs to re-hedge his position once a credit in the portfolio is replaced by another credit, the settlement of substitution has to be done in a mark-to-market way. The simplest way is to price the specific tranche with the replaced credit first, and then with the replacing credit, the difference being the substitution cost or benefit. This approach requires both parties of the transaction to agree on the model used and also the parameters used in the model. This has become possible since the Normal copula model has been accepted as the market standard for CDO pricing. The two parties usually agree on the pricing parameters, with a typical reference point being the current implied correlation embedded in the corresponding STCDO tranche. The cost of the substitution may be positive or negative — that is, paid by the tranche buyer to the dealer or from the dealer to the tranche buyer. The cost can be paid through any of the following:

1 Upfront payment between the protection buyer and the protection seller.

2 The alteration of the subordination level below the single-tranche CDO.

3 Resetting the coupon payable to the single-tranche CDO up to the point where the coupon has been reduced to zero, followed by either step (1) or step (2).

Each of the above methods of substitution settlement aims to offset the change in the mark-to-market value of the tranche before and after the substitution of the credit. The net effect on mark-to-market value of the credit substitution on one side and the cash payment, alternation of the tranche subordination, or change of the coupon on the other side should be zero. This mark-to-market change in tranche value can be captured by the net marginal credit factor.

The Marginal Credit Factors

The marginal credit factor (MCF) for each name and for each tranche measures the impact on the tranche value when the credit changes from risky status to risk-free status. It is in contrast to “jump-to-default,” and may be called the “jump-to-paradise” risk measure. This “jump-to-paradise” risk measure allows us to compare two credits in a portfolio context. The difference of MCF between the replacing credit and replaced credit is the substitution cost. That difference — net MCF — is equal to the change in the mark-to-market value of the tranche as valued with the old and new reference portfolio of credits.

MCFs tend to decrease when trading into tighter names because the value of protection provided by tranche decreases as the spread tightens. In this case, substitution results in the protection seller owing the protection buyer. MCFs also depend on the correlation of the name with the rest of the credit portfolio. Trading into lower correlated names increases the value of the protection provided by the equity tranche (and some lower mezzanine tranches), but decreases the value of protection by the senior tranche.

Approximate Substitution Cost Calculation

The problem with MCFs is that there is a need to have an agreement between dealer and investor on model, model parameters and also other model inputs, such as those single-name default swap spreads that are not easily verifiable. This lack of transparency leaves tranche investors dependent on dealer’s interpretation of risk and model parameterization. A simple way to ballpark the effect of substitution is to

Settlement of credit substitution costs is done in a market-to-

market way as a difference in tranche

values before and after the credit replacement.

The marginal credit factor can be used as

an estimate of substitution costs.



consider the net impact of the substitution on the expected loss of the portfolio. We define the impact of substitution as the difference between the expected loss of the replaced credit and the replacing credit. Based on that calculation, we adjust the subordination level of each tranche to offset the net effect of substitution on the portfolio expected loss. Difference in expected loss of two individual credits can be calculated using Bloomberg CDSW page. For example, if we would like to replace a credit A with spread of x basis points by a credit B with spread of y basis points, then we calculate the difference in the present value of expected loss of credit A and credit B as a percentage of notional amount. After that, we reduce the subordination level of by that calculated percentage difference divided by the number of underlying credits in the portfolio. This approximation is based on the assumption that the net change of expected loss after substitution is deemed as if it is the actual loss incurred by the portfolio right away. Therefore, to make the adjustment for this net loss change, we lower the size of the equity tranche by this loss change, and make the same parallel adjustments in the subordination level of all other tranches on the capital structure.

This measurement ignores the difference in spread sensitivity of each tranche as well as the credit risk on the premium side. It usually underestimates the compensation required for the credit substitution. But the exact size and direction of the difference between the approximation approach and the mark-to-market approach depend on the various parameters, such as spread levels of the replaced credits, size and subordination of the tranche and the overall spread levels in the portfolio. But the advantage of this approximation approach is that investors can easily assess the substitution cost (in terms of substitution-level change) independently from a portfolio model output.

Single-Tranche CDO Market Trading and liquidity in single-tranche CDOs has risen dramatically in recent months, since they have proven to be powerful and flexible structured credit products. The reason for their popularity is that investors can use them to execute a variety of customized investment objectives and solutions. Although these trading strategies are generally structured around the liquid market in iBoxx tranches, investors can apply the same strategies to customized portfolios in which they select the reference obligations of the single-tranche CDOs. As a combined approach, investors can start with standard iBoxx portfolio and substitute certain number of handpicked names for which they have preference not to have them included in their STCDO portfolio. As a representative example of single-tranche CDO, we present the liquid September 09 iBoxx tranches (see Figure 127). Each tranche trades at different implied correlation, which is market-driven factor in a standard copula model reflecting the demand and supply for the tranche protection. Implied correlation skew may also be due to copula model specifications and assumptions. We will explore correlation skew and its implications more fully in future research.

Trading and liquidity in single-tranche CDOs

has risen dramatically because they have

proven to be powerful and flexible structured

credit products.



Figure 127. iBoxx Tranches, Each Tranche Notional Size is $10 Million, 12 Apr 04 Tranche Maturity Implied Corr. (%) Mid Spread (bp) Credit 01 ($000) Default 01 ($000) Correlation 01 ($000)15%–30% Sep 09 28.7 10.25 2.9 4.9 5.510%–15% Sep 09 22.9 43.5 10.7 28.1 15.37%–10% Sep 09 19.6 107.5 23.2 86.0 22.83%–7%b Sep 09 — 295 45.5 282.9 0.70%–3%a Sep 09 19.8 39% 65.4 1,395.9 -81.8iBoxx b Sep 09 — 54.5 4.9 47.5 0.0a 0%–3% iBoxx tranche is quoted in points of upfront payment plus a fixed 500bp running premium. b Tranche is correlation insensitive. Credit 01 is the mark-to-market change in dollar value of a tranche if each of the names in the portfolio widens by 1bp. Default 01 is the mark-to-market change in the value of a tranche if one of credits defaults (defaulted credit is chosen to have the largest impact on the tranche). Correlation 01 is the mark-to market change in the value of the tranche if correlation changes for 1%. Risk figures are based on the implied correlation for each tranche. Source: Citigroup.

Investment Strategies Investors can use single-tranche CDOs to structure various investment strategies to target their risk/return profiles or hedging needs. We classify the strategies as:

1 Leverage strategies: Enhancing yield through leverage provided by tranching.

2 Market view strategies: Expressing a long or short view on the market using the appropriate tranches.

3 Correlation strategies: Expressing a view on implied correlation (equivalently, expressing a view on tranche technicals)

4 Relative value strategies: Taking a view on the relative cheapness of old versus new iBoxx tranches, five-year versus ten-year iBoxx tranche spreads, etc.

5 Micro/macro hedging strategies: Hedging individual default (micro) risk and/or market spread (macro) risk.

Leverage Strategies

Figure 128. Single-Tranche CDOs — Leverage Strategies Strategy Sample Trade Motivation Investors Long position (sell protection) in mezzanine or senior tranche of iBoxx or customized portfolio.

Investor sells protection on $10 million Sep 09 7%–10% iBoxx tranche.

Tranche spread (mid): 107.5bp

iBoxx spread (mid): 54.5

Date: 12 Apr 04

Investors who look for a buy-and-hold investment can obtain enhanced yield through leverage and express a bullish view.

Portfolio managers, CDO investors.

Source: Citigroup.

Long Position in Mezzanine or Senior Tranche

Investors can take a long buy-and-hold position (sell protection) in mezzanine or senior tranches on iBoxx or in customized portfolios. This trade should appeal to investors who are looking for enhanced yield through leverage. The motivation for this bullish trade is to obtain a targeted degree of leverage on the underlying portfolio while picking up spread respect to the iBoxx index or comparable corporate investments. Such buy-and-hold positions are often executed in an unfunded way, although investors can buy the credit through a funded credit-linked note, collateralized by triple-A rated paper.

A long position in a mezzanine or senior

tranche allows investors to obtain

leverage and express a bullish view.



Market View Strategies

Single-tranches CDOs that are referenced on iBoxx index products allow investors to separate their views on credit fundamentals (defaults) from their views on market and liquidity risk (spreads) and allow investors to take directional investment decision positions on defaults and spreads separately. Single-tranche trades are be structured to be superior in terms of carry, leverage, and convexity to outright long or short market positions on the iBoxx index or a portfolio of individual names. We analyzed and compared several of such trades in our recent publication, Bull and Bear in a Boxx98.

Figure 129. Single-Tranche CDOs – Market View Strategies Strategy Sample Trade Motivation Investors Short position (buy protection) in senior 7%–10% iBoxx tranche.

Investor buys protection on $10 million Sep 09 7%–10% iBoxx tranche.


iBoxx spread (mid): 54.5bp

Date: 12 Apr 04

Investors who want to protect against spread backup can execute this leveraged, low-carry, positively convex position.

Portfolio managers, bearish investors that worry about a reversal in the corporate spread rally.

Long position (sell protection) in equity 0%–3% iBoxx tranche.


Tranche spread (mid): 39 points upfront + 500bp running


Date: 12 Apr 04

Investors that are bullish about default rates and want to enter into a leveraged, positively convex, high-carry position.

Credit-savvy hedge funds.

Carry- or delta-neutral combination of short position (buy protection) in 7%–10% iBoxx tranche and long position (sell protection) in 0%–3% iBoxx tranche.

Delta-neutral trade:

Investor buys protection on $28.2 million Sep 09 7%–10% iBoxx tranche. Investor sells protection on $10 million Sep 09 0%–3% iBoxx tranche.

7%–10% tranche spread (mid): 107.5

0%–3% tranche spread (mid): 39 points upfront + 500bps running


Date: 12 Apr 04

Investors that are bearish on spreads and bullish on credit quality can structure carry-neutral or delta-neutral position and maintain positive convexity.

Hedge funds and investors expressing view on global macro risk.

Source: Citigroup.

Short Position in Senior Tranche

Investors that are bearish on overall market-spread movements driven by global factors can take a short position (buy protection) in the senior iBoxx tranche. For example, we recommend buying protection on the 7%–10% iBoxx tranche because this tranche requires a lower premium for protection per unit of spread risk than junior tranches or an outright short with the iBoxx index. This position offers approximately five times the leverage and positive convexity relative to a short on the iBoxx index, and it provides a low carry protection against the spread backup.

98 Bull and Bear in a Boxx — Using Tranched Products to Express Credit View, Arvind Rajan, Graham Murphy, and Jure Skarabot, Citigroup, February 19, 2004.

Using single-tranche CDOs, investors can

separate their views on default and spread risk.



Long Position in Equity Tranche

Investors that are bullish on the credit quality of a portfolio of names or about default rates risk in the market as a whole, can go long (sell protection) the 0%-3% tranche. This trade offers bullish investors a leveraged and positively convex position with a high carry. Of course, this trade is exposed to defaults and to spread widening. Depending on their fundamental view, investors can hedge certain individual credits in the reference portfolio by buying default protection on those names and reducing the idiosyncratic risk of the trade. After partially hedging against defaults, investors still should be left with a significant positive carry on the trade.

Carry- or Delta-Neutral Combinations of Short Senior Tranche and Long Equity Tranche

Investors that are bearish on the possibility spread selloff and bullish on credit quality can combine a long position in the equity tranche with a short position in the senior tranche to pay for the negative carry. This carry-neutral trade captures the effect of positive convexity for senior and junior tranche and generates gains in the case of larger spread moves. A short senior and long equity tranche position can also be combined into a delta-neutral trade. Compared with directional trades, this delta-neutral combination generates lower, but still positive gains in spread selloff and spread rally scenarios.

Correlations Strategies

Using a single-tranche synthetic CDO, investors can express their view on the implied correlation of tranches as different tranches can have opposite reactions to changes in correlation. Investors can choose the tranche subordination and size in line with their view on correlation. Note that the implied correlation is primarily a market-based factor, driven by the demand and supply of protection for each individual tranche. In most cases, correlation traders delta hedge their positions against the spread risk. Name-specific spread risk (micro spread risk) can be delta hedged with the single-name credit default swaps, and market spread risk (macro spread risk) can be hedged by the appropriate position in the iBoxx index. In addition, correlation traders can also hedge a portion of the default risk with a set of selected single-name credit default swaps. Therefore, these trades allow correlation traders to take a view on market demand and supply for single-tranche CDOs, without taking on credit risk.

Market view strategies can be structured as

carry or duration neutral trades, maintaining positive convexity.

Correlation traders can take a view on market

supply and demand for single-tranche CDOs.



Figure 130. Single-Tranche CDOs – Correlation Strategies Strategy Sample Trade Motivation Investors Long Correlation Trade: Short position (buy protection) in senior tranche or long position (sell protection) in equity tranche. Delta hedge with single-name CDS or iBoxx index.

Investor buys protection on $10 million Sep 09 7%–10% iBoxx tranche. Investor sells protection on $47.3 million Sep 09 iBoxx IG index



Date: 12 Apr 04

Investors who want to have long correlation exposure and expect implied correlation to increase. Positive carry.

Correlation traders (hedge funds, bank proprietary desks, and broker/dealers) with a positive view on implied correlation.

Short Correlation Trade: Long position (sell protection) in senior tranche or short position (buy protection) in equity tranche. Delta hedge with single-name CDS or iBoxx index.

Investor sells protection on $10 million Sep 09 7%–10% iBoxx tranche. Investor buys protection on $47.3 million Sep 09 iBoxx IG index.



Date: 12 Apr 04

Investors who want to have short correlation exposure and expect implied correlation to decrease.

Correlation trades (hedge funds, bank proprietary desks, and broker/dealers) with a negative view on implied correlation.

Correlation-Insensitive Trade: Long (sell protection) or short (buy protection) position in junior mezzanine tranche.

Investor sells or buys protection on $10 million Sep 09 3%–7% iBoxx tranche.

Tranche spread (mid): 295


Date: 12 Apr 04

Investors that are uncertain about correlation levels and don’t want to take correlation risk, but want to invest in single-tranche CDOs to get leverage or take a directional market view position.

Investors who don’t want to take exposure to changes in correlation.

Source: Citigroup.

Short Position in Senior Tranche or Long Position In Equity Tranche (Long Correlation)

Investors, who prefer to be long correlation, can set up such a position by going short the senior tranche (buying protection) or going long the equity tranche (selling protection). A rise in senior tranche implied correlation would imply that increased demand for protection is pushing senior tranche spreads wider and, all else equal, making the trade profitable. A rise in equity tranche implied correlation would imply that increased supply of protection is pushing equity tranche spreads tighter and all else equal, making the trade profitable.

Long Position in Senior Tranche or Short Position in Equity Tranche (Short Correlation)

Investors can choose a short correlation position by going long the senior tranche (selling protection) or going short the equity tranche (buying protection). A decrease in senior tranche implied correlation would imply that the senior tranche is trading at lower spreads, making the trade profitable, with all else equal. A decrease in equity tranche implied correlation would imply that increased demand for protection is pushing equity tranche spreads wider, making the trade profitable.

Long or Short Position in Junior Mezzanine Tranche (Correlation Insensitive)

If investors are uncertain about default correlation levels, and they do not wish to take correlation risk, they can invest (sell protection) in the tranche on the capital structure that is correlation insensitive. Typically, this tranche would be the junior mezzanine (second loss tranche), but the exact attachment point and size depends on the characteristics of the collateral pool. Proactive investors can achieve a similar correlation-neutral position by investing in a more junior and a more senior tranche below and above the correlation inflection point. These two tranches have opposite sensitivities to the correlation, and investors can earn a higher spread with this position than investing in the correlation insensitive mezzanine tranche alone.

A junior mezzanine tranche is usually

correlation insensitive, suitable for investors

who do not want to take on the correlation risk.



Although at origination the specific tranche can be correlation insensitive, with time-decay the correlation sensitivity will change.

Relative Value Strategies

Figure 131. Single-Tranche CDOs – Relative Value Strategies Strategy Sample Trade Motivation Investors Long/short position in on-the-run iBoxx tranches versus short/long position in matching off-the run iBoxx tranches.

Investor buys protection on $10 million Sep 09 7%–10% iBoxx tranche. Investor sells protection on $10 million Mar 09 7%–10% iBoxx tranche.

Sep 09 7%–10% tranche spread (mid): 107.5bp

Mar 09 7%–10% tranche spread (mid): 112.5bp

iBoxx Sep 09 spread (mid): 54.5bp

iBoxx Mar 09 spread (mid): 56bp

Date: 12 Apr 04

Relative value trade opportunities at the same leverage level between the new Sep 09 iBoxx tranche and the old Mar 09 iBoxx tranche. Investors can focus on implied correlation as relative value measure.

Hedge funds, correlation traders, broker/dealers.

Long/short position in five-year iBoxx tranches versus long/short position in matching ten-year iBoxx tranches.

Investor sells protection on Sep 09 7%–10% iBoxx tranche. Investor buys protection on Sep 14 7%–10% iBoxx tranche.

Relative value trade opportunities between the five-year and ten-year iBoxx tranches. Investors use implied correlation skew surface as relative value measure.

Hedge funds, correlation traders, broker/dealers.

Long/short position in 3%–100% iBoxx tranches and short/long position in iBoxx IG index versus short/long position in 0%–3% iBoxx tranche.

Investor buys protection on $10 million Sep 09 3%–100% iBoxx tranche. Investor delta-hedges by selling protection on $6.8 million Sep 09 iBoxx IG index.

Tranche spread (mid): 20


Date: 12 Apr 04

Relative value trade opportunities between the traded 0%–3% iBoxx tranche and synthetically replicated equity piece using 3%–100% iBoxx tranche and iBoxx index.

Relative value hedge funds.

Source: Citigroup.

Long/Short Position in On-The-Run iBoxx Tranches Versus Short/Long Position in Matching Off-The-Run iBoxx Tranches

The recent roll of the iBoxx indexes provided a set of relative value trade opportunities in the single-tranche market. With the two-sided market in on-the-run and off-the-run iBoxx tranches, investors can now compare matching tranches referenced to similar underlying portfolios. Investors can focus on the implied correlation of new and old iBoxx single-tranches and look for relative value opportunities at the same leverage level in the capital structure.

Long/Short Position in Five-Year IBoxx Tranches Versus Short/Long Position in Ten-Year iBoxx Tranches

Expecting that liquid market for ten-year iBoxx tranches will improve further, we believe that investors will be able to execute relative value trade strategies based on the comparison between the five-year and ten-year iBoxx tranches referenced to the same portfolio. Implied correlation skew surface should be an indicative cheap/rich measure for individual tranches.

On-the-run and off-the-run iBoxx tranches

provide potential relative value opportunities.



Long/Short Position 3%–100% iBoxx Tranche With Short/Long Position in iBoxx Index Versus Short/Long Position in 0%–3% iBoxx Tranche

Investors can combine a 3%–100% iBoxx tranche and delta-hedge it with the iBoxx index to synthetically replicate the equity 0%-3% tranche of the iBoxx. Comparing the obtained spread premium with the market spread for the traded equity 0%-3% equity tranche of the iBoxx allows investors to enter into a relative value trade or execute the directional view by using a cheaper way to access the equity part on the iBoxx capital structure.

Micro/Macro Hedging Strategies

Figure 132. Single-Tranche CDOs – Micro/Macro Hedging Strategies Strategy Sample Trade Motivation Investors Short position (buy protection) in equity tranche on a customized portfolio of names.

Investor buys protection on 0%-3% tranche of a customized portfolio. Investor sells protection to the same counterparty on individual CDS names (“exchange the deltas”).

Cheaper way to hedge against default risk than buying protection on each individual name.

Portfolio managers and bank loan portfolios.

Short position (buy protection) in equity tranche on a customized portfolio of names and long position (sell protection) in mezzanine iBoxx tranche.

Investor buys protection on 0%–3% tranche of a customized portfolio. Investor sells protection on Sep 09 7%–10% iBoxx tranche

Portfolio managers with a positive view on macro risk can compensate the cost of hedging by going long senior iBoxx tranche.

Portfolio managers or buy-and-hold investors that are not subject to mark-to-market accounting treatment.

Short position (buy protection) in senior iBoxx tranche.

Investor buys protection on $10 million Sep 09 7%–10% iBoxx tranche.



Date: 12 Apr 04

Investors that already hedged the most risky names against default and want to hedge against market-wide spread widening. This strategy is more efficient hedge than a short iBoxx position.

Portfolio managers or buy-and-hold investors with strong fundamental approach to portfolio selection.

Long position (sell protection) in senior iBoxx tranche.




Date: 12 Apr 04

Portfolio managers that have bought default protection in single-name CDS market and want to protect their hedging portfolio against mark-to-market risk.

Bank loan portfolio managers.

Source: Citigroup.

Short Position in Equity Tranche

A short position (buying protection) in an equity tranche on a customized portfolio of names is an efficient way to hedge against default risk. Investors select names from their portfolio that, in their view, have high risk of default and the protection on the first-loss tranche of this portfolio. Usually, investors sell protection to the same counterparty on individual CDS names (“exchange the deltas”). Buying protection on the equity tranche of a customized portfolio is a cheaper way to hedge against default than buying protection on each individual name. In addition, investors that are comfortable with a certain number of defaults being unhedged can buy protection on the second-loss tranche. The higher attachment point of the tranche can substantially lower the cost of carry.

A short position in an equity tranche can be

used as a hedge against defaults (micro hedging).



Short Position in Equity Tranche and Long Position in Mezzanine Tranche

Hedging against default using individual credit default swaps or a short position in the equity tranche can be an expensive strategy. Investors that are less sensitive to market spread widening or that have a positive view on the general macro risk factors but worry about defaults in their portfolio can compensate for their hedging cost by going long a senior tranche on the iBoxx. The premium received on the senior tranche protection should lower the cost of protection on the equity tranche.

Short Position in Senior Tranche

Certain investors are searching for solutions that will protect them against general market spread widening in their portfolios, especially if they have already hedged the most risky names against default or they trust in their selection of credits based on their fundamental views. For such investors, hedging with the iBoxx is not the most efficient solution, because the outright market short does not separate default and spread risk protection. A more effective hedging strategy is to put on a short position (buy protection) in the senior tranche that has the highest Credit 01-to-carry ratio. We estimated that the 7%–10% iBoxx tranche provides investors with the most suitable hedge, after taking into account the liquidity component of traded iBoxx tranches.

Long Position in Mezzanine or Senior Tranche

A long position in mezzanine and senior iBoxx tranches is a leveraged position on market-wide spread movements. If investors, such as bank loan portfolio managers, have bought protection in the single-name CDS market against defaults in their portfolios, this hedging portfolio is exposed to mark-to-market risk driven primarily by macro factors. If investors want to protect against mark-to-market risk in their hedging portfolios, then a long position in the mezzanine or senior iBoxx tranche can provide the solution. Because the junior tranches take on relatively more default risk than senior tranches, they are much less suitable for such hedging purposes. In addition, a long position in the iBoxx index is a less efficient hedge against market-wide spread moves, because the index spread is affected as by the systematic market risk as by the credit-specific events.

Future Trends and Conclusion Single-tranche CDOs are one of the important innovations among structured portfolio credit products. Extending their features and flexibility above the standard cash and synthetic CDO products, they allow investors to place various leveraged positions in the credit market and efficiently execute their investment strategies. Flexibility in choosing the reference portfolio allows investors to structure the portfolio investments that can be properly matched with their investment portfolios, as a separate investment strategy or as a hedging position.

It is our expectation that the single-tranche CDO products will remain an established component of the structured credit market. New products based on single-tranche methodology will become available, driven by the investors’ focus on the alternative asset classes and the correlation between them. As one of the main trends we expect further standardization of single-tranche structures. As the default index linked tranches will trade under the same documentation, investors will be able to unwind their positions with any of the market makers. At the same pace, we expect that the derivative market on the standard single-tranches will develop to the mature stage.

A short position in a senior tranche can be

used as a hedge against market-wide spread

moves (macro hedging).



Trading volume in the swaptions on single tranches referenced to the default index will provide the liquidity and opportunity to buy or sell volatility positions. We also expect that the forward market on tranched indices will expand, providing investors with further opportunities to express directional views.

We expect that the demand for single-tranche CDO will further increase with the growth of CDO of CDOs transactions. Single-tranche CDOs are flexible and suitable source of collateral. Currently, such transactions (CDO squared) already include single-tranche CDOs in their collateral pool. In search for yield, future CDO squared structures will source even larger pool of single-tranche CDOs.

The single-tranche methodology and structures are not unique to the liquid names in the investment-grade corporate universe. They can also be applied to other reference classes, from high-yield reference obligations to the whole variety of the structured finance products. Following the increased liquidity in high-yield and emerging markets default indexes (e.g., high-yield and emerging market iBoxx default indexes) and variety of other asset classes, we expect that in the future single-tranche CDOs will reference not only the corporate names, but a whole range of the market segments. These new product will provide investors with an efficient tool to structure their investments according to their views on macro trends and correlation between asset classes.

FEBRUARY 19, 2004

Arvind Rajan Graham Murphy Jure Skarabot

Bull and Bear in a Boxx — Using Tranched Product to Express Credit Views



Su

mm

ary

Despite improving credit fundamentals and declining defaults,

investors worry that historically tight spreads may prove vulnerable

to interest rate volatility and the timing of a change in monetary

policy. We show how tranched credit products allow investors to

separate their view on credit fundamentals (defaults) from their view

on liquidity (spreads). Buying protection on senior tranches

expresses more of a bearish view on spreads than on defaults. We

show it to be cheaper, more liquid, and more positively convex

position than an outright credit short. Selling protection on junior

tranches expresses a bullish view on credit fundamentals (defaults).

Each trade has advantages over outright long or short positions

using bonds or a single-name default swap index.

We also recommend several combinations, expressing a variety of bullish and bearish views, based on buying protection on the 7%–10% tranche and selling protection on the 0%–3% tranche of the iBoxx investment-grade default index (iBoxx CDX.NA.IG). The trades can be structured as a carry-neutral combination, which is still bearish on the market spread, or as a duration-neutral trade with respect to the market spread. Alternatively, an investor with a more bullish view on the credit quality can sell more protection on the junior tranche and further profit from spread tightening. Each trade combination offers superior performance in certain scenarios compared to an outright short or long on the underlying portfolio.

The out-of-the-money nature of buying protection on senior tranches makes it positively convex. Rising spreads bring it closer to being at-the-money, making it more sensitive to spreads as spreads rise. Selling equity protection is also positively convex for similar reasons. An investor buying senior protection or selling equity tranche protection is long implied correlation, which has declined recently. Implied correlations could rise if a Fed hike or other macro risks rise.

Although these strategies offer many advantages, several caveats relating to market liquidity, bid-ask spreads, model risk, and other factors should be considered before embarking on these trades.



Bull and Bear in a Boxx — Using Tranched Product to Express Credit Views

Credit Fundamentals and Spreads — Destined to Diverge? There is good reason to believe that global liquidity is at least as important as fundamental credit expectations in driving credit spreads. Today this liquidity takes the form of low yields and accommodative monetary policy. Fixed-income investors’ search for yield has caused a cascade down the spectrum of credit, contributing to the 84bp rally99 in corporate investment-grade spreads in 2003.

The key dilemma for credit investors today may be in taking advantage of improving credit fundamentals while avoiding buying into too-tight credit spreads near the peak of the liquidity cycle. In the current cycle, markets have been wary of interest rates rising and the Fed since last June, when, partly in response to FOMC language, the ten-year spiked from 3.1% to 4.5%, and spreads and equities sold off in sympathy. Although rates and spreads subsequently recovered, the heightened concern over interest rates is ever-present. This can be seen in Figure 133, which shows that the correlation between corporate spreads and Treasury yields — historically negative most of the time — has turned positive over the past several months.

The positive co-movement is not surprising because the generally negative historical correlation between rate and spread moves over the past decade reversed during Fed tightening cycles. Our recent study of eight Fed tightening cycles since 1971 shows that spreads moved wider by an average 44bp during tightening cycles,100 especially in the latter half of the cycle. Other analysis also indicates widening at least in some credit spreads101 when the Fed futures curve first prices in a Fed hike.

99

“A New Year in the Corporate Bond Market,” Adler Dennis, et al., Citigroup, January 26, 2004.

100 For a study of eight tightening cycles since 1971 see “The Effect of Fed Tightening on Corporate Bond Spreads,” Terry

Benzschawel, Bond Market Roundup: Strategy, Citigroup, December 12, 2003.

101 This effect has been shown to be significant in a panel analysis of emerging market spreads.



Figure 133. Ten-Year Treasury Yields Versus Correlation of Treasury Yields and Corporate Spread Moves, 1990–2004

0

1

2

3

4

5

6

7

8

9

10

Jan 90 Jan 92 Jan 94 Jan 96 Jan 98 Jan 00 Jan 02 Jan 04

10-Y

ear T

reas

ury

Yiel

d (%

)

(0.8)

(0.6)

(0.4)

(0.2)

0.0

0.2

0.4

0.6

0.8

12-M

onth

Rol

ling

Corr

elat

ion

10 Yr Treasury Yield

Correlation of Treasury Yields and Corporate Spreads Moves

Source: Citigroup.

The Fed’s most recent change in language in its statement in January — when it removed the phrase saying that it would stay on hold for a “considerable period” — came as the first shock to financial markets in 2004. If history is any guide, it will not be the last. Investment-grade spreads, which widened from 47bp to 61bp on the investment-grade iBoxx102 in response, have since rallied back to 56bp. We expect that 2004 will bring several such episodes of selloff and tightening, and spreads will be especially vulnerable to interest rate volatility.

In the meantime, the bullish signals on credit are compelling. US economic growth (expected to be 5.2% in 2004 and 4.3% in 2005) has been blistering. From a historical point of view, default rates and ratings downgrades tend to track long-term economic cycles closely, with previous peaks in 1990–1991 and 2002. There are already strong indications of a continued improvement in corporate balance sheets; data on corporate leverage shows stabilization in total debt to total capital and a rise in pretax interest coverage, and a decline in total debt to earnings. In Figure 134, we show that the rate of investment-grade corporate blowups has collapsed from its peak in fall 2002 along with the level of corporate spreads. Corporations seem to have learned their lesson from 2002, using historically tight yields to term-out debt instead of increasing leverage, thus gradually reducing their vulnerability to a future credit crunch.

102 For the purpose of this report, iBoxx index refers the North American investment-grade iBoxx default index (CDX.NA.IG).



Figure 134. Historical Blow-up Rates and Investment-Grade Credit Spreads, 1990–2004

0

1

2

3

4

5

6

7

1Q 90 2Q 91 3Q 92 4Q 93 1Q 95 2Q 96 3Q 97 4Q 98 1Q 00 2Q 01 3Q 02 4Q 03Quarter

Blow

up R

ate

(%)

50

100

150

200

250

300

Mea

n Qu

arte

rly O

AS (b

p)

Average Monthly Blowup Rate

Average OAS

Mean = 0.6; Median = 0.2

Source: Citigroup.

The question is: Which of these forces will prevail when the Fed is ultimately forced to raise interest rates? Many investors believe that even though tighter monetary policy may indeed be a response to continued strength in the economy, and the fundamentals of credit may continue to improve, the spread rally, constrained in part by historically low rates, could reverse. Getting fundamentals (credit and defaults) right, but spreads wrong, is not a luxury investors can afford, especially if they mark-to-market. And even if the Fed tightening is a long way off, last month’s volatility may be the first of many false alarms in 2004 that increase volatility and erode the profitability of already tight spread-based carry products.

An Outright Short Is an Expensive Answer

To restate the dilemma, a long credit position, whether through a bond or a single-name CDS, combines spread market risk and name-specific credit risk. So, if spreads are likely to move the same way as Treasuries during a rate-hike cycle, a credit investment with a sound fundamental underpinning nonetheless becomes an implicit wager on interest rates. The obvious way to protect against market-wide spread widening is to hedge long positions or get short credit spreads outright.

Yet this approach has several major drawbacks. First, shorting investment-grade credit103 is expensive (about 56bp per year of carry over LIBOR; about 100bp over Treasuries). Investors must be very bearish on spreads to be willing to incur this much loss of carry.104 Second, shorting through specific corporate bonds or single name CDS exposes investors to name-specific (upside) risk on those names and offers limited liquidity. Investors may be unwilling to pay to take such risks. Third, an outright short on credit throws the baby (reduced default risk) out with the bathwater (increased spread widening risk), forcing investors to forego long positions in improving stories. Fourth, short bond (or long single name CDS

103 Investment-grade credit is represented by the investment-grade iBoxx index. 104 To illustrate, assume a carry of 75bp over risk-free rates, a duration of 5.0 for fixed-income investments, and further assume that spreads stay the same with probability 30%, up 50bp with probability p% and down 50bp with probability 70-p% by year-end. Then, based on a simple break-even analysis between the expected loss in a back-up and the carry, an investor must believe that a spread back up is 2.5 times as likely as a rally in order to want to forego the 75bp carry all year long.



protection) positions are negatively convex105 — they become progressively less profitable as spreads widen.

A Structured Solution to the Paradox

These considerations have kept many traditional credit investors from implementing bearish views, explaining the apparently puzzling paradox that many credit investors seem to believe that spreads are unsustainably tight, and yet most are unwilling to short the credit market or to hedge spread risk. We believe that such investors should take a look at the recently introduced synthetic portfolio products, particularly credit index products in tranched form. In this note, we will show that buying protection on a senior tranche of an index such as iBoxx provides leveraged, low-carry, positively convex protection against a spread backup.

Further, investors who are bullish about credit fundamentals should sell protection on an equity tranche of names they like. This position offsets the negative carry of the long protection position and offers less systematic market risk and more credit-specific exposure than owning the same names outright. A customized portfolio may be substituted for the index on either leg of the trade. The rest of this report explains how and why each leg of the strategy works, and also how one may combine the two into a neutral or positive carry position that reduces a long portfolio’s exposure to macro market risk.

The Bottom Line — Recommended Trades In Figure 135 we outline the trades we would recommend to investors who are either bearish on market spreads or bullish on credit fundamentals (default risk) of the market or of specific credits. We also show how to appropriately combine these trades to create a bearish, market-neutral or bullish combination. For purposes of comparison, we also present the P&L of short and long credit positions in the iBoxx index.

Figure 135. The iBoxx Index and Tranches, Trade Scenarios, Three-Month Mark-to-Marketa P&L, 11 Feb 04

Amount ($MM) Three-Month Three-Month Mark-to-Market ($000)

Strategy Buy Protection Sell Protection Carry ($000) -25bp -10bp Unch +10bp +25bp One DefaultBearish Spread (tranche) Short 7–10% 5.95 (14.25) (212) (110) (15) 102 306 30Bearish through iBoxx Short iBoxx portfolio 10.00 (14.25) (106) (44) 0 44 110 49

Bullish Credit (tranche) Long 0–3% 1.14 14.25 232 88 12 (58) (144) (137)Bullish through iBoxx Long iBoxx portfolio 10.00 14.25 106 44 (0) (44) (110) (49)

Combo — Carry Neutral Short 0–3% and long 7–10% 5.95 1.14 — 20 (22) (4) 44 161 (107)Combo — PV01 Neutral Short 0–3% and long 7–10% 5.95 1.95 10.15 185 40 5 3 59 (204)Combo — PV01 Bullish Short 0–3% and long 7–10% 5.95 2.28 14.25 252 65 8 (13) 17 (244)a These mark-to-market figures assume constant implied correlation for each tranche over the duration of the trade. Note that changes in implied correlation and bid-ask spreads will impact the profitability of these trades. Spread scenarios also assume no defaults. Spread changes are simultaneous changes to all underlying namesb Three-month P&L equal to sum of three-month carry and three-month mark-to-market. Recovery assumption: 40%. Source: Citigroup.

105 The negative convexity is just the reverse of the positive convexity of fixed-rate bonds, which in turn results from the fact that because of the convex shape of the price-yield curve, a bond’s price drops less and increases more than its duration times the yield increase or decrease. Similarly, buying protection, which is equivalent to shorting a bond, is negatively convex.



Bearish Spread Trade

As we will see, senior tranches are more sensitive to spread changes and relatively less sensitive to defaults. So for investors bearish on market spread movements driven by global factors such as rates, we recommend buying protection on the 7%–10% tranche (“Bearish Spread” trade). This trade is superior to an outright short on the reference iBoxx per unit of negative carry paid. This position offers both leverage (greater sensitivity to spreads) and positive convexity relative to a short on the credit index (represented by iBoxx). For the same level of carry, in a scenario of spread selloff, going short credit with the 7%-10% iBoxx index tranche generates two to three times higher mark-to-market gains than a short position in the iBoxx index. The short position in the senior tranche is also positively convex with the gain in the case of a 25bp spread selloff being 1.5 times the loss in 25bp spread rally. Therefore, its performance is superior to iBoxx in terms of the upside/downside ratios corresponding to symmetric 25bp spread moves.

Bullish Credit Trade

For investors bullish on the credit quality of a portfolios of names, or about default rates in the market as a whole, we suggest selling protection on the 0%-3% tranche (“Bullish Spread” trade). This trade offers a leveraged and positively convex long on credit outcomes. For the same amount of carry, investors can have twice the gain by going long credit on the 0%-3% tranche versus a long position in the iBoxx index. Similar to a short position in the senior tranches, a long position in the equity tranche is also positively convex, with one and a half times higher gain in case of a spread rally versus a loss in spread selloff corresponding to a symmetric 25bp spread moves. Of course, this trade is exposed to defaults and to spread widening.

Combination Trades

It is possible to combine these trades to advantage in a number of ways. In “Combo — Zero Carry” we use a long credit position in the junior tranche to pay for the negative carry of the short credit position in the senior tranche. This carry-neutral trade captures the effect of positive convexity for senior and equity tranches and it generates gains in both scenarios corresponding to symmetric 25bp upside and downside spread moves.

It is also possible to combine the trades into a bullish stance while hedging against market spread selloffs. We show two versions of this. The first is duration neutral (“Combo — PV01 Neutral” trade). We show that this trade position has positive mark-to-market gains in cases of smaller (e.g., 10bp) and larger (e.g., 25bp) symmetric spread moves. Comparing to the directional trades, duration-neutral combination generates lower, but positive gains in spread selloff and spread rally scenarios. Alternatively, if we’re more bullish on credit quality than bearish on spread selloff, we can sell more credit protection on the junior tranche in the above duration-neutral combination. This brings in a greater carry and profits from spread tightening (“Combo — PV01 Bullish” trade) while limiting losses in most market selloffs. In this trade combination, investors gain two and a half times more than going long the iBoxx Reference Index if spreads rally, while curtailing spread-selloff-driven losses.



The reader will note that combining long and short credit positions in tranches allows superior performances in a variety of bullish and bearish scenarios, except outright default. How is this possible? In the remainder of this report we explain the intuition behind the leverage and convexity and the trade-off between spread and default risk that result in these scenario returns.

Introducing Credit Indexes and Their Tranches Over the past year, a more liquid market in single-name credit default swap (CDS) has allowed credit indexes such as iBoxx CDX and Trac-X to be formed. These are equal-weighted indexes of five-year CDS that trade daily in the market and allow investors to buy or sell a broad credit basket with instant depth and liquidity. The investment-grade iBoxx index, which trades with a 0.5bp to 1bp bid-ask spread under normal conditions, consists of 125 equal par-weighted CDS.

Definitions Let us introduce some basic terms. The “bid” is the annualized premium in basis points (bp) that the protection buyer would pay to the seller of protection to compensate for potential losses on a tranche. To bond investors, it is equivalent to the spread paid over LIBOR for a risky credit. The “ask” is the annualized premium that the dealer would require from the buyer of protection. The PV01 is the mark-to-market change in the dollar value of a tranche if each of the names in the portfolio widens by 1bp. (Reported PV01s generally assume a $10 million notional amount.) Thus the PV01 is a good measure of the sensitivity of a tranche to broad spread market moves. The “default 01” is the mark-to-market change in the value of tranche if one of the credits defaults (defaulted credit is chosen to have the largest impact on the tranche). It is therefore a good measure of the sensitivity of the tranche to name-specific risk. The “carry” is the premium times the tranche size, or the annual dollar amount received/paid for loss protection on the tranche, assuming no defaults.

The PV01 is similar to spread DV01 of a single bond or CDS in that it denotes the price sensitivity to the spread. Note, however, that traditional fixed-income intuition can be misleading here. Due to increased leverage, the junior tranches have larger PV01 numbers.106 Note however that the higher PV01s denote greater price sensitivity to spread but not a longer average life, as they usually do for bonds. In fact, the risk-neutral weighted average life of the cash flows is between 4.5 years and 5 years for most of the tranches except the equity tranche, for which it is 2.5 years to 3 years because defaults curtail some of the cash flows in that case.

Figure 136. The iBoxx CDX IG Index and its Tranches, 11 Feb 04 Tranche Maturity Notional ($MM) Bid (bp) Offer (bp) PV 01($000) Default 01($000)15%–30% Mar 09 10 8.5 16.5 3 510%–15% Mar 09 10 46 55 11 297%–10% Mar 09 10 89.5 102 21 833%–7% Mar 09 10 285 345 43 2820%–3%a Mar 09 10 41.8% 43.75% 63 1,357iBoxx Index Mar 09 10 56.5 57.5 5 48a 0%–3% iBoxx CDX tranche is quoted in points upfront payment plus a fixed 500bp running premium. Source: Citigroup.

106 When spreads are high enough, however, the junior-most tranches may in fact have lower PV01s than some intermediate tranches.



The market innovation that allows us to weight default and spread risk differently is tranching, whereby the cash flows from the iBoxx are further divided into buckets, typically labeled the “x%–y%” tranche. We provide a listing of liquidly traded tranches with some basic data in Figure 136. There is implicitly a “30%-100%” “super-senior” tranche, which does not trade generically, but which can be constructed synthetically from the others and whose spread would theoretically be close to zero. The key feature of the tranches is that they absorb credit risk sequentially, like a CDO. The first 3% in losses resulting from default are absorbed by the 0%–3% tranche, the next 4% of losses (if necessary) are absorbed by the 3%–7% tranche, and so on.

Points Up Front

Generally, the 0%–3% tranche (frequently referred to as the equity or first-loss tranche) generally trades with “points up front” and a fixed premium 500bp annually. This is equivalent to a risky bond (in particular, a LIBOR+500bp floater) that trades on a dollar-price basis. The usage of points as the trading convention reflects the higher level of default risk in the equity tranche.

Liquidity, Leverage, and Efficiency The iBoxx tranches constitute a CDO-like structure that trades transparently and relatively liquidly in the market. The approximate market bid-offer spreads in Figure 136 indicate there is liquidity in these tranches, though less than in the iBoxx index itself. Note that the bid-ask spread on these tranches varies both by seniority and that the most senior tranche that trades in the inter-dealer market, the 15%–30% tranche, is less liquid then the 7%-10% tranche.

The standardized tranches represent a major improvement in the transparency of the tranched market. Traded volumes are rising steadily, reaching $500–$600 million per week in January 2004 of traded standardized tranches on the iBoxx reference portfolio. The two-way markets in these securities indicate to investors the degree of pricing efficiency and mitigate concerns about fair pricing. Further, this type of tranching can be done in principle on any portfolio. Naturally, customized portfolios may entail higher transaction costs and offer less liquidity compared to generic index-based tranches that already trade in the market.

Figure 137. Default and Spread Sensitivity and Efficiency, 11 Feb 04

Sensitivity Efficiency

Mid Spread (bp) One-Year Carry ($000)a PV 01($000) Default 01 ($000) PV01/Carry Default 01/Carry15%–30% 12.5 13 3 5 0.24 0.4310%–15% 50.5 51 11 29 0.21 0.587%–10% 95.75 96 21 83 0.22 0.863%–7% 315 315 43 282 0.14 0.890%–3%b 1871 1,871 63 1,357 0.03 0.73iBoxx Index 57 57 5 48 0.08 0.83a Assuming no defaults in a one-year period. Based on notional $10,000,000 for each tranche. b Annualized spread (bp) equivalent to points upfront plus 500bp running premium. Source: Citigroup.



Leverage — A Byproduct of the Seniority Structure

In Figure 137 we illustrate a basic feature of iBoxx tranches, namely the leverage they provide. This leverage is a function of two basic factors, the subordination level and the size of the tranche. This leverage is manifested in the much higher premium or carry of the more junior tranches. The increased premium is compensation for the greater risks carried by the junior tranches, and the PV01 and default 01 are the two most important risk measures. While the PV01 of all the tranches approximately add up to that of the iBoxx as a whole,107 which has a PV01 of $4,646 on a $10 million tranche notional, the junior tranches are much more sensitive than the senior tranches. For example, the 7%–10% tranche has a PV01 of $20,806 (on a $10 million tranche notional), which is 4.5 times that of the iBoxx, and the 0%–3% and 3%–7% tranches have similar high PV01s (about 9.2 and 13.6 times that of the iBoxx). The sensitivity of the 15%-30% senior tranche is less than that of the portfolio.

Since tranches absorb losses sequentially, the default 01, which is the change in dollar value due to an individual credit loss, is even more unequally divided than the PV01, with the 0%–3% tranche having much greater share of the default 01 than the others. The default 01 drops off sharply for senior tranches. Thus the iBoxx tranches allow investors nonrecourse leverage to go long or short the market. Figure 137 shows, for each the tranches, the relative sensitivity with respect to spread and default.

The Concept of Spread Market Efficiency and Single Credit Efficiency

The carry associated with each tranche is shown in Figure 137 in dollars. As expected, the more junior the tranche, the higher the carry to compensate for higher risk. We define the “spread market efficiency” of a tranche as the amount of PV01 per unit of carry. It is a ratio that indicates the cost-effectiveness of a tranche as a vehicle to express a view on spread risk. Analogous to spread market efficiency, let us define “single credit efficiency” of a tranche as the amount of default 01 per unit of carry. Both of these measures are shown for the iBoxx tranches in Figure 137. Note that while every tranche has both spread market and single-credit default sensitivity, the two risks are distributed differently among the tranches.

Why Senior Tranched Credit Makes an Efficient Short

These sensitivity numbers reveal that the senior tranches have the highest spread market efficiency, meaning that they require a lower premium for protection per unit of spread risk than junior tranches (see Figure 137) This high efficiency lies at the heart of the argument for using senior tranches to short against spread market-wide back-up. It also shows that senior tranches are not as efficient for going long credit risk; nonetheless, their relative safety may make them appropriate for many investors. It is important here to draw the contrast to traditional leverage, for example, to buy or short on margin. Traditional leverage allows the amount of risk or sensitivity per unit of capital to be scaled up, but typically the ratio of risk to the carry (the income from a long position or the running cost of a short position) does not change. So what is

107 If implied correlations were identical for all tranches, they would add up exactly.



different here? Why is an investor able to get greater spread sensitivity per unit of carry with, say, the 7%–10% tranche than with the 0%–3% tranche?

Single Credit Efficiency

The key to understanding this is the difference in how default risk and spread risk are distributed among the tranches. The most senior tranches (such as the 15%-30% tranche) have very little of either default or spread risk. The equity and junior tranches, by construction, have most of the credit specific risk. Since carry is compensation for the both kinds of risk, it is natural that relatively more of the spread risk (per unit of carry) is carried by the middle tranches.108 There is, therefore, no free lunch here, and the greater efficiency of, say, the 7–10% tranche in helping the bearish investor to short market risk means the same tranche is less effective than junior tranches in protecting against a default in one of the underlying names. The advantage, however, is that the investor is able to hedge against interest rate-inspired volatility in spreads without also paying for a hedge against actual defaults.

Equity Tranches — An Efficient Way to Go Long Default Risk

Conversely, many credit investors specializing in fundamental name-specific analysis generally have strong views on the default likelihood and creditworthiness of credits they follow. Even after the massive rally in spreads there are many strong credit stories in today’s declining default environment. As an example, lesser-rated credits (BBB and BB) are trading historically cheaper than higher rating categories. Further, because many categories of bond investors have absolute yield bogeys, and because underlying Treasury rates are at historically low levels, there are many historically tight credits that nevertheless offer value on fundamental grounds, namely willingness and ability to pay. An investor who does not expect much spread tightening but also expects low default risk should sell protection on the 0%-3% tranche of an underlying portfolio of carefully chosen names, because this tranche offers the greatest absolute carry and also offers the lowest spread market efficiency — that is, the maximum carry per unit of spread market risk. In practice, this is often done by selling protection on the equity tranche of a generic portfolio, and then buying protection on the single names that are deemed most risky by the investor.

Tranche Convexity — Where it Comes From We had mentioned that buying protection on senior tranches offers “positive convexity.” What we mean by this is explained in Figure 138, which shows, for each of the tranches, the P&L (from the perspective of protection buyer) for bullish and bearish scenarios — a 25bp rally in underlying spreads and a 25bp backup in underlying spreads (assuming instantaneous spread moves). The convexity column shows the extra P&L over that obtained by multiplying the PV01 by the spread move. From Figure 138, it is clear that while long protection positions in the iBoxx index and 0–3% tranche are negatively convex, the higher tranches have positive convexity when used as long protection positions. Note, however, that the convexity of the iBoxx is relatively small.

108 Note that we are speaking in “per unit of carry” terms here, and that in par amount terms, a junior tranche has more of both default and spread risk.



Figure 138. Scenarios for Long Protection Positions Using iBoxx Tranches Based on Instantaneous Spread Moves, 11 Feb 04 (Dollars in Thousands)

MTM With Spreads Tightening 25bp MTM With Spreads Widening 25bp

Tranche Total ($) Duration ($) Convexity ($) Total ($) Duration ($) Convexity ($)15%–30% (48) (73) 25 89 76 1310%–15% (186) (257) 71 302 268 357%–10% (357) (499) 142 585 520 653%–7% (1,044) (1,022) (16) 1,056 1,071 (15)0%–3% (1,863) (1,523) (340) 1,371 1,586 (215)iBoxx Index (114) (112) (2) 110 116 (7)

Source: Citigroup.

We have already mentioned that simply buying protection on the market is negatively convex for all the same reasons that a long position on a bond is positively convex, (although the effect is small in this case due to the short maturity of the instruments).109 If so, why does buying protection on a senior tranche offer positive convexity? The best way to understand this is through an analogy with how simple options work. Recall that an option that is struck at the money (ATM), where the strike price is the same as the market price, has greater sensitivity to the price of the underlying security than an out-of-the-money (OTM) option. Thus, OTM options become more sensitive to the price as the market price moves closer to the strike price.

Buying protection on a senior tranche is analogous in its payoff structure to a call spread. As shown in Figure 139, the tranche loss profile with respect to total portfolio loss resembles the payoff profile at expiration of a call spread with respect to the price of the underlying. Buying protection is thus equivalent to owning an out-of-the money call spread on default-driven losses in the underlying portfolio.

Figure 139. Analogy Between Tranche Loss and Spread Option Payoff Tranche Loss Versus Portfolio Loss Call Spread Payoff

0

1

2

3

4

5

6

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Portfolio Loss

Tran

che

Loss

Tranche Loss

0

1

2

3

4

5

6

0 2 4 6 8 10 12 14Price

Call

Spre

ad P

ayof

f

Call Spread Payoff

Source: Citigroup.

109 There is another sense in which some participants call buying protection at today’s historically tight levels a “convex” trade: they are referring to the likelihood that spreads have much greater potential to go up than down since they are historically tight. But this has nothing to do with the convexity of any instrument, tranched or otherwise. It is really a comment that the distribution of future spreads is likely to be asymmetric. Thus this “convexity” refers to the asymmetry of spread levels (and therefore changes) associated with opposite states of the world, put simply, that spreads can only contract to zero in the best of worlds, but can widen a lot more if things do not go well. That is, a spread distribution is bounded by zero and has a fat tail. This fact is important to consider in modeling of default and spread risk. However, for the sake of clarity, we will never use the term “convexity” to refer to it.



Tranche convexity can be observed and quantified by plotting the increase in the expected loss on a tranche as the spreads widen (see Figure 140). Consider buying protection on a senior tranche as an out-of-the-money option on defaults on the underlying bonds in the portfolio. The first few defaults do not affect the payout of the tranche; the tranche cash flows are threatened only when the loss from defaults approaches the subordination level of the tranche. As spreads go up, defaults become more likely, so the OTM option comes closer to becoming an ATM option, and the tranche sensitivity to spreads increases. Indeed, we observe in Figure 140 that senior tranches become more sensitive to spreads as spreads rise. The effects of this positive convexity are quite large. As Figure 138 shows, positive convexity can contribute 10% to 20% to the P&L move of a 7%–10% tranche for a spread move of 25bp.

Figure 140. Effect of Spread Increase on Expected Tranche Loss

0

300,000

600,000

900,000

1,200,000

0%–3% 3%–7% 7%–10% 10%–15% 15%–30% iBoxx IndexTranche

Incr

emen

tal E

xpec

ted

Loss

+ 25 bp

+ 50 bp

+ 75 bp

+ 100 bpTranche Notional $10,000,000

Source: Citigroup.

Conversely, protection bought on a sufficiently junior tranche is either at-the money or in-the-money. Thus an increase in spread reduces its sensitivity, as it goes deeper in-the money. As Figure 140 shows, the change in expected loss on junior tranches decreases as spreads widen. The opposite trade, selling protection on an equity tranche, is like buying an out-of-the money call110 on the riskiest part of the underlying portfolio: if default risk decreases (that is, spreads tighten) the equity tranche becomes more sensitive to spreads, thus offering positive convexity to those using equity tranches to get long credit risk.

Correlation — A Proxy for Macro Risk Factors The most talked-about factor in the tranched market is the notion of correlation. We left this concept for last because investors often get hung up on it. Indeed, the effect of implied correlation on prices in correlation models is complex, and tends to cause confusion; we have explained the most widely used framework, the normal copula model in detail elsewhere.111

110 We use the call as the most intuitive analogy, but to be more accurate, the position is really akin to selling a deep-in-the-money put.

111 “A Copula Function Approach to Credit Portfolio Modeling,” Quantitative Credit Analyst, Li D., Connor J, Gu A., Citigroup, May 2003 and “The Impact of Correlation on Tranched Credit Risk,” Quantitative Credit Analyst, Ling A., Murphy G., Citigroup, May 2003.



Correlation Models

The gist of the quantitative approach is the following. Defaults in a portfolio are not independent events — they tend to cluster together. This means that the probability of a default of a single name in a portfolio in a specified period is dependent on whether or not other names default in the same period; that is, their default probabilities are correlated. In the widely used copula framework, this dependence is modeled using a joint distribution of the so-called survival functions of credits (this function specifies the likelihood of a credit not having defaulted yet as a function of time and is equivalent to specifying the default probability at each point in time).

The correlation between names, which can be thought of as asset correlations between these credits, is a key input that affects this joint distribution. For simplicity, although different pairs of names may have different default correlations, the practice is to use either a single correlation parameter across the board (or in some cases two parameters, an inter-industry and a cross-industry sector correlation).

Correlation and Tranche Value

But the key take-away from the model is very simple: the value of equity tranches increases as the correlation parameter in the model is increased, whereas that of senior tranches drops. The intuition behind this relationship is also simple. An increase in default correlations, while keeping overall risk constant, that means prospects will either get better for all credits or will go wrong at once across the board. This is in contrast to a low-correlation universe, in which defaults occur independently of one another and do not affect the rest of the market. So when the correlation parameter is high, there are many more outcomes when many defaults occur simultaneously. This is bad for investors who are long credit through senior tranches, which could suffer losses in these cases. But a junior tranche, by definition, absorbs the early losses in the portfolio. So the increased likelihood of a large cluster of defaults (beyond the expected number) may not hurt a junior tranche as much, because the maximum loss of a tranche is its full notional amount and the tranche is indifferent to losses past this amount.

By the same token, when implied correlation is high, there are also more cases of very few defaults across the board. This benefits junior tranches, particularly the equity tranche. Thus plugging in a higher correlation parameter will tend to raise the value of the equity tranche while it reduces the value of senior tranches (see Figure 141). Once individual spread levels are fixed, there is a direct mapping between the value (or spread level) of a particular tranche and the correlation parameter. This is the reason why the correlation parameter value corresponding to a particular spread level on a particular tranche is called the implied correlation of that tranche, Over time, changes in implied correlation can be inferred from the changing market prices of the various tranches (note that implied correlations have no effect on the value of the portfolio as a whole).



Figure 141. Effect of Move in Implied Correlation on the Expected Loss for Each Tranche (US Dollars in Thousands)

(600)

(400)

(200)

0

200

400

600

- 5% - 2.5% Market Level + 2.5% + 5%Tranche Implied Correlation Change

Dolla

r Cha

nge

in T

ranc

he E

xpec

ted

Loss

15%–30% 10%–15%7%–10% 3%–7%0%–3% iBoxx Index

All tranches are $10 million notional. Source: Citigroup.

Systematic Versus Credit-Specific Factors Are Key to Correlation

But another angle to understanding correlation is from the standpoint of what we have been talking about already — systemic spread market risk versus single-credit risk. The factors affecting credit spreads can be broadly classified as credit specific factors and systemic (market-wide) factors112. The former type includes capital structure, profitability, earnings stream, quality of management, investment in R&D, product pipeline, competitive landscape, etc. The latter includes the term structure of risk-free rates, valuations of competing investments, global risk appetite, the threat of catastrophic events such as war, disease and large-scale terrorism, etc.

When spreads move, a combination of both types of factors may be at work, but their effects are different. Single-credit risk is specific to a name, and its effects should not create much contagion in the market. An increase in systemic risk moves the whole market simultaneously, and over time, can be manifested as a rise in realized correlation. The effect of such systemic risk may be small, such as a small rise in global interest rates, or large, such as the risk of a large scale terrorist attack or war. In turn, over time, realized correlation changes can affect implied correlation. A change in implied correlation affects the way that expected losses are distributed through the capital structure. This can ultimately lead to a selloff in senior tranches.113 Note that this effect of increased correlation would occur in addition to the effect of higher spreads in the underlying names.

112

A third category includes market inefficiencies such as costs of credit analysis, non-transparency, lack of liquidity, etc.

113 This claim will need to be validated empirically as we gather pricing history for implied correlation.



Figure 142. iBoxx CDX Tranches, Implied Correlation, Oct 03–Feb 04

0%

5%

10%

15%

20%

25%

30%

35%

40%

29 Oct 03 26 Nov 03 24 Dec 03 21 Jan 04 18 Feb 04

0%–3% 3%–7% 7%–10%10%–15% 15%–30%

Source: Citigroup.

Buying Senior Protection Protects Against Rising Correlation

In practice, implied correlation is hugely affected by the supply and demand dynamics for individual tranches. Over the past few months, implied correlation has been stable, and on a mildly declining trend (Figure 142). A long protection position in senior tranches is thus a view that systematic credit risk, having been on a long decline, could increase again as rates rise or the Fed moves closer to tightening. Note that this exposure to correlation is distinct from, and in addition to, the sensitivity to a market spread widening. Similarly, since equity positions are positively affected by a rise in correlation, selling protection in an equity tranche also benefits from increased correlation. Of course, a further drop in correlation would negatively affect both positions.

The recent rise in positive correlation between rates and credit spreads, and the episodic selloffs in June–July last year and again last month both suggest that volatility in rates is among the key universal factors affecting credit risk correlation. However, we would warn that despite its plausibility, this notion remains a hypothesis to be tested in the absence of empirical evidence. For this reason, in our scenario analysis of P&L, we assume that implied correlation stays the same. If we increased the correlation in the bearish scenarios (and/or reduced in bullish scenarios for spread), it would tend to make our recommendations to sell protection on junior tranches and buy on senior tranches perform even better.

Putting it All Together — Tranche Returns Under Bullish and Bearish Scenarios What impact does the combination of leverage, efficiency, and correlation have on the returns of the tranches compared to the iBoxx index. In Figure 143, we provide a P&L analysis for each of the tranches and the index, from the perspective of the protection seller, based on scenarios where, over a period of three months, spreads remain unchanged, rally by 10bp or 25bp or sell off by 10bp or 25bp. We also indicate what occurs if there is one default during the three month period. Note that carry P&L



(premium income) has been separated from mark-to-market P&L for clarity of exposition. The total P&L in each scenario (shown in Figure 143 for some combinations of the trades below in the same scenarios) is the sum of the two components.

Figure 143. The iBoxx Index, Scenarios for Long Credit Positions Based on Three-Month Mark-to-Market P&L, 11 Feb 04

Three-Month Mark-to-Market P&L ($000)a

Tranche Annualized Carry ($000 ) -25bp -10bp Unch +10bp +25bp 1 Default15%–30% 13 48 25 3 (25) (78) (1)10%–15% 51 184 95 12 (89) (266) (15)7%–10% 96 356 185 26 (172) (514) (50)3%–7% 315 1,035 442 45 (368) (969) (230)0%–3%b 500 2,035 770 102 (506) (1,266) (1,200)iBoxx Index 57 106 44 (0) (44) (110) (49)a Based on notional of $10,000,000 on each tranche. Assuming no defaults in the three-month period. Assuming constant implied correlation for each tranche. b 0%–3% iBoxx CDX tranche is quoted in points upfront payment plus a fixed 500bp running premium. Source: Citigroup.

Looking at the carry and no-change columns, we note that the 7%–10% tranche has almost double the carry of the iBoxx. The positive convexity of the 7%–10% and the other two senior tranches is manifested by the asymmetry of mark-to-market returns between the +25bp and -25bp scenarios. However, in the case of a default, the iBoxx loses less per unit of carry than the 7%–10% tranche. The 10%–15% tranche is not very sensitive to default and seems to offer comparable carry to the iBoxx. But in a spread sell-off of +25bp, it loses 2.5 times as much as the iBoxx. However, although the 10%–15% tranche seems quite attractive as a hedge for spread risk, its lack of liquidity does not allow it to be used as a practical trading instrument. It may be used only as a long-term buy and hold instrument expressing specific views on spread and default. For this reason, we have not included it in our bearish recommendations. The 15%–30% tranche similarly does not trade much.

In the unchanged scenario, the 0%–3% tranche offers the greatest carry relative to exposure (the loss in the bearish scenarios), while the positive convexity of the 0%–3% tranche as a long-credit instrument shows up in the higher returns in bullish scenarios compared to losses in the bearish scenarios. Thus the 0%–3% is clearly superior to the iBoxx in expressing a positive credit view while shielding against market volatility. However, the 0%–3% has immediate exposure to default and hence loses twice as much as the iBoxx in the event of a default relative to its carry. Being highly leveraged inherently, the 0%–3% tranche also stands to lose its entire principal (notional) if a sufficient number of losses due to default occur.

We also observe in Figure 143 (see “unchanged spreads scenario”) that in the first three months, the tranches have positive change in value with respect to time decay. The reason for this effect is that the premium leg decay is linear with time and the expected loss leg is non-linear, decreasing faster at the beginning and slowing towards the end of tranche life for senior tranches. Note that the opposite holds for the equity tranche; the expected loss decay is slower at the beginning and faster towards the end of tranche life. But because the protection on the equity tranche is paid as points upfront plus lower running premium, the mark-to-market change in value over the first three months is actually positive, as the points upfront payment decays at the same rate as the expected loss on the equity tranche.



The ideas and calculations discussed here helped us to select instruments and calculate projected P&L for the scenarios in Figure 135. We hope they have helped you to understand where the combination of convexity and leverage comes from, and to demystify one of the most powerful instruments available today to position for various credit market outcomes.

Some Observations and Caveats ➤ Unwinding of trades may occur at different implied correlations from inception

because of bid-offer spreads and changes in market levels. The difference in liquidity between the iBoxx and the tranches will therefore hurt the profitability of tranched positions.

➤ The iBoxx index is reconstituted every six months, with the next roll date due on March 20, 2004, and hence positions after that date may become less liquid. The extent of name substitution is likely to be limited, so the old iBoxx index and tranches, although likely to become less liquid, are likely to remain tradable after that date. Nevertheless, investors intending to hold the trade past the switchover date may in some cases be better off waiting until after the new index has begun trading. This issue does not apply to investors looking to put on the recommended trades on custom portfolios.

➤ Sensitivities and P&L using constant implied correlation are model dependent — although the copula model is the standard, actual changes in the values of tranches may vary from projected changes.

➤ Observations and results are only valid at the current level of market and credit spreads. For example, behavior of senior tranches will be more like that of junior tranches if market spreads are at higher levels than those used in the calculations.

➤ The sensitivity to individual credits depends on the individual spread levels and weights in portfolio. For example, tranches are more sensitive to spread widening or default of low-spread credits than high-spread credits.

➤ Actual market moves will not result in equal moves in all credits, hence spread PV01 (which is based on moving all underlying credits by 1bp) is a theoretical concept that cannot be used in predicting P&L accurately for actual market moves.

➤ Liquidity conditions can change. The participation in iBoxx, and particularly in its tranches, is still limited to a narrow segment of market participants. Although we currently expect the market to improve in breadth and depth, it is still very new. It is possible, as with other recently established markets, that trading volumes and liquidity can reverse in adverse market conditions.

FEBRUARY 19, 2004

John Carpenter Milena Todorova Pradeep Kumar Urvish Bidkar

Trading the CDX.EM iBoxx The New CDS Benchmark for Emerging Markets



Su

mm

ary

The New Emerging Markets Benchmark ➤ The CDX.EM iBoxx is a new tradable index on 14 emerging market credits

from Latin America, CEEMEA, and Asia. It is unfunded and consists of a basket of five-year credit default swaps. As such, it has far less Treasury duration than the ESBI and, thus, offers a purer view of emerging market credit.

Standardized, Liquid, and Efficient ➤ Portfolio construction is standardized and transparent. Rebalancing occurs

every six months according to published rules and dealer inputs on country weights. Eleven dealers have committed to making markets. The bid-offer spread has been approximately $0.25 (6bp–8bp) with weekly notional volume of $200–250 million. We expect volume to increase further.

Applications: Diversified Long/Short Exposure, Relative Value, and Portfolio Rebalancing ➤ We believe that the CDX.EM iBoxx is useful for taking broad long/short

exposure to a diversified pool of emerging market credits, expressing a number of different relative value ideas, and aiding in portfolio rebalancing of large exposures.

Statistical Relationships with Individual Countries and the ESBI114 ➤ We reconstruct a history of hypothetical CDX.EM iBoxx returns and

discuss statistical relationships with component countries. We calculate hedge ratios of the CDX.EM iBoxx versus individual names and the ESBI.

Average Credit Rating is BB+ ➤ The average credit rating in the CDX.EM iBoxx (weighted according to

iBoxx weights) is BB+. Currently the CDX.EM iBoxx is 2bp wide to the BB corporate iBoxx.

Replicating the iBoxx with Bonds ➤ We show that CDX.EM iBoxx returns can be well replicated with as few as

five countries, by decomposing the index according to credit rating. One investment-grade country behaves as a proxy for all the similarly rated credits in the index. Best results are obtained using eight countries that have the highest original CDX.EM iBoxx weights with disproportionately higher weightings of Korea, Malaysia, or Poland.

114 The average life of the hypothetical CDX.EM iBoxx that we construct will not precisely match the average life of the actual iBoxx CDX.EM.



Trading the CDX.EM iBoxx

Introduction On January 20, the CDX.EM iBoxx — a new emerging market credit default swap benchmark index — began trading. The initial spread of this index was set at 265bp and began trading at 99.30–99.60. This index is composed of 14 emerging market sovereign reference entities spanning Latin America, CEEMEA, and Asia (see Figure 144). Buying an CDX.EM iBoxx contract is similar to selling default protection on 14 different countries with notional amounts according to the weights in Figure 144. Since its inception, daily volume has ranged between $50–75 million and the bid/offer spread is typically $0.25 (6bp–8bp). In this article we highlight the main features of this index and applications.

Since the CDX.EM iBoxx index began trading, it has widened 51bp, to 316bp, while the ESBI cash index has widened 10bp to a yield of 6.74% (the spread over Treasuries of the ESBI has widened 30bp from 312bp to 342bp).

Figure 144. CDX.EM Composition, Reference Bonds,a Mid-Market CDS Levels, and Bid/Ask Spreads, 19 Mar 04

Country iBoxx Weight (%) Reference SecurityMid Five-Year

CDS Spread (bp) CDS Bid-Ask SpreadBrazil 14 12.250% 3/06/30 553 30Russia 14 5.000% 3/31/30 195 14Mexico 13 7.500% 4/08/33 116 10Turkey 11 11.875% 1/15/30 374 40Colombia 8 10.375% 1/28/33 410 40Venezuela 8 9.250% 9/15/27 685 70Korea 7 8.250% 1/15/14 57 10Philippines 6 8.250% 1/15/14 515 20Malaysia 4 7.500% 6/01/13 50 10Bulgaria 3 8.250% 1/15/15 120 20Panama 3 8.875% 9/30/27 280 40Peru 3 8.750% 11/21/33 400 50Poland 3 6.250% 7/03/12 33 10South Africa 3 8.500% 6/23/17 110 30a A triggering default event on a particular credit in CDX.EM iBoxx is defined as a missed payment on the reference bonds listed. Source: Citigroup.

CDX.EM iBoxx Mechanics — How Does it Work? The iBoxx will be rebalanced twice a year and trades with standard ISDA documentation. The CDX.EM iBoxx trades on a price basis, making mark-to-market profit and loss easy and transparent. As each contract ages, the biannual issuance ensures that investors have the opportunity to roll into the new and liquid on-the run contract. Alternatively, closing transactions can be easily accomplished because of the standardized contract maturities and spreads.

The current on-the-run CDX.EM iBoxx expires on June 20, 2009. The initial contract spread is the median of the estimated contract spread values submitted by members of the CDX.EM iBoxx consortium. With the exception of the first contract, which began trading early on January 20, 2004, a new five-year contract will begin trading on March 20 and September 20 of each year. Payments occur on June 20 and



December 20 with a one-time upfront payment to equate the fixed leg to the current market value (this payment could flow from buyer to seller or vice versa depending on which way the credit has moved since the index was last set).

The fixed leg of the contract underlying the index pays on June 20 and December 20 for 11 total payments (the tenor varies from 4.75 years to 5.25 years depending on when the contract was bought). The first coupon is always short. The very first contract, introduced on January 21, 2004, and expiring on June 20, 2009, is slightly longer than usual (in the future it will typically start trading in March). The first roll occurs on September 20 of this year into a December 2009 contract.

On March 20/September 20, the index “rolls” into a contract of 5.25 years with (possibly) rebalanced country weights and a new fixed spread. The weights can be changed according to dealer input because of changes in the liquidity of an underlying credit. The new fixed spread reflects the average of prices submitted by various dealers, and thus the contract should initially trade near 100.

The price is quoted clean as a percentage of par notional (for example, 97). The difference between the price and 100 (for example 100 – 97 = 3) represents the present value of the difference between the fixed leg specified in the contract (currently 265bp) and the current market value. A one-time upfront payment (3% of notional, for example) is made to equate the two.

Example of Trading

Let us suppose that on March 11, 2004, the market maker quotes an offer price of 97.0 to an investor who wishes to take a diversified exposure (long) to emerging market credits by buying CDX.EM iBoxx (that is, selling protection to the dealer). The coupon equivalent spread on the current CDX.EM iBoxx is 265bp. On a $10 million notional, the cash flows are as follows:

Figure 145. CDX.EM iBoxx Cash Flow Example

Upfront payment on March 11, 2004, of10,000,000 * (100 - 97.00) / 100 =$300,000 - accrued

Eleven payments of(0.0265/2) * 10,000,000 = $132,500each, one on each Jun 20/Dec 20through June 2009.

Investor Dealer

Source: Citigroup.

What Happens in a Credit Event?

In our hypothetical scenario, suppose that a reference entity, which we will call reference entity A, defaults in May 2004, and reference entity A has a weighting of 10% in the index. The current CDX.EM iBoxx contract is divided into two contracts with notionals of $1,000,000 and $9,000,000, respectively. The protection seller in CDX.EM iBoxx delivers $1,000,000 cash to the protection buyer and, in return, receives physical delivery of $1,000,000 face value of deliverable obligations of reference credit A. After the credit event, the current CDX.EM iBoxx continues to trade with weights normalized to have the same proportions but without the defaulted



entity. A contract held prior to default would be equivalent to $9,000,000 worth of notional on the contract trading after default. That is, if you bought CDX.EM iBoxx (sold protection) and A defaults, you would be left holding a bond of A with $1,000,000 face value, $9,000,000 notional of the current CDX.EM iBoxx (which no longer includes A), and would have paid out $1,000,000 in cash.

Figure 146. Example Cash Flows When an CDX.EM iBoxx Constituent Country Defaults

$1,000,000 cash

$1,000,000 face value of eligiblebonds issued by A.

Investor Dealer

Source: Citigroup.

Converting Price to Spread

CDX.EM iBoxx is quoted in price. To translate the conversion formula to spread, imagine first that a CDS contract is quoted in spread terms. Suppose an investor buys protection at 200bp and, following a market move, sells the exact same contract for 250bp. The investor has locked in a risky payment of 50bp. The mark-to-market of this contract is the present value of the 50bp cash flows, discounted at the risky rate that is implied by current CDS prices. Working backwards, the spread on a current CDX.EM iBoxx is the value x at which a stream of payments of amount x (discounted at the current risky rate) equals the upfront payment that is quoted from the dealer. What follows is an example:

CDX.EM price is quoted at 97.00. The contract pays 265bp;

DF(i): Discount factor from time i implied by the LIBOR swap curve;

RDF(i) : Risky discount factor to allow for default probability;

x : the current spread adjustment that must be added to the 265bp;

Since the contract makes 11 semiannual payments, where i is in years:

)2

()2

(11

1 200.397100.100

iRDF

iDF

i

xEMCDX ∑

===−=−

Because x does not depend on i,

∑=

= 11

1)

2()

2(

0.6

i

iRDF

iDF

x

This equation can be solved for x, which is the only unknown. If x was 75bp, then the equivalent contract would be 265bp + 75bp = 340bp. So, which risky curve should be used? Although it would be possible to build a flat default probability curve from the currently quoted CDX.EM price, CDSs on individual countries almost always show an upward sloping default probability term structure. We recommend that all available prices on CDS contracts on the underlying credits be used to build a risky curve for each credit using reasonable recovery values in default (25% would be



typical). The cumulative risky discount curve can then be aggregated as the weighted sum of the individual discount factors in proportion to their notional CDX.EM weights. As a quick approximation, one could also simply divide the price difference by the appropriately weighted sum of the DV01s of a five-year CDS of each underlying credit.

This conversion can be done with Bloomberg analytics by typing CDX CDS <CORP> <GO>, selecting “EM” and typing <DES> <GO> to use the CDSW function. By adjusting the contract spread to make the contract value close to 0, the equivalent spread difference can be obtained.

Useful Applications of CDX.EM iBoxx

An Efficient Way to Get Long or Short Exposure to Diversified Emerging Market Credits

The CDX.EM iBoxx provides an efficient and cost effective way to take a diversified view on the overall market. Selling or buying protection on a large number of credits individually incurs high transaction costs. This index has less treasury duration than a cash index, and thus offers a purer view of credit. As an alternative to investing in a cash index, CDX.EM iBoxx allows investors to receive extra carry through the typically positive default-cash basis. Investors prohibited from investing in unfunded transactions, but wanting a diversified exposure, would be able to buy a credit-linked note tied to the performance of CDX.EM iBoxx. Dealers would be able to charge a smaller bid-offer spread on the credit linked notes because it will be easy for them to use the CDX.EM iBoxx instead of shorting individual securities or credit default swaps.

Hedge a Portfolio of Cash Bonds Against a General Selloff

Buying protection via the CDX.EM iBoxx allows investors to reduce their credit exposure without selling their cash instruments. This short is especially useful given that cash instruments can be illiquid and securities can become hard to borrow in the repo market. In addition, the potential for short squeezes during crises remains high.

As shown in Figure 147, the spread changes of the CDX.EM iBoxx basket are highly correlated to the EM proxy index, which in turn are highly correlated to the cash index (ESBI).115 Note that the slope of the line in Figure 147, which is an estimate of the beta of spread changes of CDX.EM iBoxx to the emerging market index, can be used in calculating the amount of CDX.EM iBoxx needed to hedge the index. In Figure 147 the beta is 1.15, suggesting that a lower notional on the CDX.EM iBoxx is needed to hedge overall exposure to cash index.

115 Here, the EM proxy index consists of iBoxx countries weighted by their market values in the ESBI Index. iBoxx proxy spreads are calculated by weighting the ESBI spread of each iBoxx country with its weight in the iBoxx index.



Figure 147. Changes in CDX.EM iBoxx Proxy Versus Changes in EM Sovereign Proxy, Mar 02–Mar 04

y = 1.1534x + 2.0196

R2 = 0.9591

-250

-200

-150

-100

-50

0

50

100

150

200

250

-200 -150 -100 -50 0 50 100 150 200

EM Sovereign (ESBI) Proxy Monthly Spread Changes (bp)

EM iB

oxx

Prox

y M

onth

ly S

prea

d Ch

ange

s (b

p)

Source: Citigroup.

Express a View on the Relative Performance of Individual Countries Versus the Overall EM Market

By going long an individual credit (or pool of credits) and shorting the CDX.EM iBoxx, investors can take a view of relative outperformance against overall emerging market credit. This trade may be attractive to investors who do not have an overall market view or wish to isolate credit specific risks from systematic ones. In such trades, the choice of hedge ratio can depend on a number of factors including the beta between the country and index, the duration of the country specific instrument, carry considerations, and default risk.

The Positive Correlation Between Individual Credits and CDX.EM iBoxx Can Also Be Useful in Portfolio Rebalancing

If an investor wanted to liquidate a large position in a particular bond over the course of days or weeks, he could first sell CDX.EM iBoxx (that is, buy protection) and then gradually buy back his CDX.EM iBoxx as he sold the bonds. This would allow the investor to immediately tap the liquidity of CDX.EM iBoxx and thereby protect against a general emerging market selloff that could occur before he adjusts his bond position.

Figure 148 shows the beta and correlation between monthly returns of individual countries versus the index. Because idiosyncratic events can make these parameters unstable over time, we calculate them over different windows dating back one, two, and four years from the present. These results show the following:

➤ Beta and correlation vary dramatically across credits, even similarly rated ones such as Brazil and Colombia.

➤ Many countries have reasonably stable betas over time.

Brazil and Turkey have high betas and behave this way consistently over both long-and short-term horizons. Both are highly correlated with the index. Among the higher rated credits, Mexico has a higher beta than Poland, South Africa, Bulgaria, and Chile. Finally, betas for Turkey, Venezuela, and the Philippines have been increasing over time, while Russia and Brazil’s have been decreasing.



Figure 148. Correlations and Betas of Individual Countries Versus CDX.EM iBoxx Proxy

One-Year Two-Year Four-Year

Beta Correlation (%) Beta Correlation (%) Beta Correlation (%)Korea 0.28 69 0.04 25 0.12 53Mexico 0.40 61 0.57 92 0.57 89Brazil 1.98 86 2.86 94 2.11 86Venezuela 1.72 80 1.07 76 0.82 69Russia 0.47 66 0.67 82 1.41 70Turkey 2.15 68 1.30 78 1.33 74Philippines 1.02 73 0.42 62 0.63 70Colombia 0.96 68 1.14 80 0.98 76Panama 0.61 46 0.71 83 0.71 81Peru 0.76 53 1.12 87 1.01 71South Africa 0.26 37 0.29 66 0.41 64Bulgaria 0.07 15 0.37 77 0.62 74Poland 0.19 56 0.16 72 0.11 54Malaysia 0.37 66 0.12 46 0.24 59Chile 0.19 73 0.27 73 0.26 66

Source: Citigroup.

Relative Value Between Emerging Market Sovereigns and High Grade/High Yield Corporates

Given the recent development of iBoxx indices of US corporate credits such as the investment-grade (CDX.IG), single-B rated (CDX.B), double-B rated (CDX.BB), and high yield (CDX.HY), it is now possible to express a view on the relative cheapness of emerging market credit spreads versus corporate spreads. Eight out of 14 credits in the CDX.EM iBoxx are rated double B or lower, carrying an allocation of 56%. Measuring the rating of each credit in the CDX.EM iBoxx by its iBoxx weight gives an average rating of BB+. The CDX.EM iBoxx index, currently at 316bp, is approximately 2bp wide to the CDX.BB iBoxx index. Those who think emerging markets are cheap to double-B corporates could buy the CDX.EM iBoxx and short the CDX.BB iBoxx.

How to Replicate the CDX.EM iBoxx With a Subset of Countries It is a matter of interest to know whether the CDX.EM iBoxx spread return (over the Treasury component) can be reasonably copied with a subset of the countries that are part of CDX.EM iBoxx.

Portfolio 4, the portfolio consisting of the subset of countries with the highest CDX.EM iBoxx weights, gives the lowest tracking error between the mimicking portfolio and CDX.EM iBoxx country returns. The weights assigned to this subset are a bit different from the original CDX.EM iBoxx weights on these countries. Korea, in particular, receives a disproportionately higher weight of 17% relative to its CDX.EM iBoxx weight of 7% (see Portfolio 4 in Figure 149). We believe that the relative overweight of Korea absorbs the allocation to a “low-risk” investment in the replicating portfolio, to the extent that the high credit quality of Korea is a good proxy for a low-risk credit that helps reduce the return tracking error of the portfolio relative to the CDX.EM iBoxx spread return.



Figure 149. Weights of Subsets of Countries to Help Replicate CDX.EM iBoxx Proxy Spread Returns CDX.EM iBoxx Country iBoxx Weight (%) Portfolio 1 (%) Portfolio 2 (%) Portfolio 3 (%) Portfolio 4 (%) Portfolio 5 (%) Portfolio 6 (%)Mexico 13 41 35 30 16 18 17Brazil 14 16 15 12 15 14 14Venezuela 8 9 8 8 9 8

Russia 14 13 13 14 15 14 15Turkey 11 12 11 12 11 11 12Philippines 6 18 16 13 8 6 11

Colombia 8 11 10 11 9Panama 3 Peru 3

South Africa 3 Bulgaria 3 Poland 3 14

Malaysia 4 17 Korea 7 17

Tracking Error (Annual) 1.75 1.46 1.04 0.61 0.68 0.68

Source: Citigroup.

The CDX.EM iBoxx proxy spread return is constructed as a weighted sum of the ESBI country spread returns according to CDX.EM iBoxx weights. The goal is to generate as closely as possible a history of how the CDX.EM iBoxx would have traded had it been in existence. Each country spread return is computed as the total return less the Treasury return on a portfolio of bonds whose duration matches the country duration for the respective country in the ESBI. The weights applied to arrive at the index return are the CDX.EM iBoxx country weights. An optimization procedure is performed to derive the optimal set of weights given a subset of CDX.EM iBoxx countries. The optimization minimizes the squared sum of monthly tracking errors over a sample of five years between the subset portfolio spread return and the return on CDX.EM iBoxx.

Forming a portfolio of the eight countries with the highest representation in the index gives the smallest error. To put errors into the context of return variability, we estimate the annualized tracking error, which is the standard deviation of the difference in annual returns between the full iBoxx CDX.EM and the replicating portfolio (see Figure 149, last row). For example, if CDX.EM iBoxx had had an annual return of 8%, Portfolio 1, with its 1.75% tracking error, would have been in the range of 6.25%–9.75% (8% +/- 1.75%) within the confidence level of 1 standard deviation. We can see an indication as to why this happens from the results of the optimization procedure when the subset weights are nonbindingly restrained to be less than or equal to one. This version of the optimization leaves about 13%–15% of weight unassigned to any country when Korea is excluded from the subset. In other words, to minimize tracking error, about 15% are set aside for a “riskless” investment.

To the extent that Korea, Malaysia (Portfolio 4), and Poland (Portfolio 5) are high-grade credits, including them and imposing a binding constraint for the weights to add up exactly to one makes the least-risky countries pick up the slack weight. Korea is a slightly more preferable choice than Malaysia and Poland as Portfolio 4 — which holds the eight countries of highest original representation in the CDX.EM iBoxx — produces the smallest tracking spread return error.



Summary CDX.EM iBoxx is a tradable index on 14 emerging market countries. We expect continued grow in its popularity because of its liquidity, efficiency, and standardization. The credit default swap technology employed in the construction allows leverage, easy short exposure, and insulation from treasury risk. Uses include: hedging against overall EM contagion, expressing relative value with component countries or related asset classes, and in portfolio rebalancing.

We have reconstructed a hypothetical CDX.EM iBoxx history and calculated its statistics with component countries and a proxy to the ESBI. We show how to replicate the CDX.EM iBoxx with a small number of bonds.

FEBRUARY 19, 2004

Jure Skarabot Richard B. Salditt Glen Taksler Dennis Adler

New Roll of the iBoxx CDX Indexes



New Roll of the iBoxx CDX Indexes

New Roll of the iBoxx CDX Indexes The iBoxx credit default indexes roll to a new set of credits on March 20, 2004, with the first day of trading on Monday, March 22, 2004. In this section, we review the mechanics of the roll, analyze the changes in the composition of the indexes, and discuss implications for investors. In evaluating the roll, investors should pay particular attention to two things: (1) the improvement in credit quality resulting from additions and deletions to the new indexes and (2) the six month maturity mismatch between the old and new indexes.

Mechanism of the Roll for Investment Grade iBoxx CDX Indexes

The iBoxx indexes roll every six months, on March 20 and September 20. To ensure objectivity and transparency, the consortium of dealers that designed the index have designated a third party (iBoxx) to be responsible for the administration of the removal of names from the current index, and coordinates determination of the new index at each roll-date. The process works as follows:

1 One week before each roll-date, each of the consortium members submits (a) a list of entities that have been downgraded below investment grade by either S&P or Moody’s, (b) a list of entities for which a merger or corporate action has occurred that makes the issuer unsuitable for the inclusion in the index, and (c) a list of entities for which CDS contracts have become significantly less liquid.

2 The administrator removes from the index all issuers that have been downgraded below investment grade by either S&P or Moody’s. The administrator then removes entities that received a majority vote for deletion with respect to (b) and with respect to (c).

3 To replace deleted entities, the consortium members now submit a list of new entities that they want added to the index (members submit twice the number of entities required after the elimination round). The administrator adds to the new index those entities that receive the highest number of votes, until the index reaches 125 names. After the names are made public, consortium members submit votes for each fixed-rate coupon (in increments of 5bp), and the median of submitted rates becomes the new coupon.

As the first roll-date is approaching, we would like to provide investors with an analysis of roll-related changes and their effect.

Changes in the iBoxx CDX Indexes

Investment Grade

This week the dealers’ consortium has submitted their votes for the replacement and substitution of the names in the index. Based on the voting outcome, six entities have been eliminated from the IG iBoxx index. All of these names have been eliminated by the rollover rules following the downgrade below the investment grade, except AT&T Wireless, which has been bought by Cingular (CNG)116 (Cingular is already in

116 Citigroup Global Markets is an advisor to BLS in Cingular's proposed merger with AT&T Wireless.



the IG iBoxx index). Six new entities will replace the old ones in the new iBoxx index (see Figure 150).

Figure 150. Deletions and Additions of the Entities in the iBoxx Index (CDX.NA.IG ), (Spreads in bp, Mid Levels), 18 Mar 04 Deletions Spread Replacements SpreadVisteon Corporation (VC) 251 Kerr Mcgee Corp. (KMG) 81Sun Microsystems, Inc. (SUNW) 145 Intelsat (INTEL) 85Toys R US Inc. (TOY) 263 CenturyTel Inc. (CTL) 105Computer Associates (CA) 134 Supervalu INC. (SVU) 72Amerada Hess Corp (AHC) 93 CVS Corp. (CVS) 43AT&T Wireless Services Inc. (AWE) 73 Alltel Corp. (AT) 53

Average 160 Average 73

Source: Citigroup.

The dealer consortium probably tried to minimize the potential disruption following the roll and has selected only the names that have been forced out of the index by the recent rating actions or mergers. Furthermore, although not required by the rules, the selection of the new names in the index almost matched the sector specifications of the eliminated names, with the exception of the industrial sector index (CDX.NA.IG.IND), which will have now one name less after the elimination of Visteon, and the consumer sector index (CDX.NA.IG.CONS), which will now have one name more after the elimination of Toys R Us and addition of CVS and Supervalu.

High Volatility iBoxx

High volatility iBoxx index (CDX.NA.IG.HVOL) went through a more substantial rearrangement, as the dealers repeated the voting on the high vol index. The new high vol iBoxx index has been selected by the voting mechanism where each of the dealers in the consortium submitted a list of 40 names from the new iBoxx composite portfolio. After the vote, 30 names with the highest score have been selected as the new high vol iBoxx index. As a result of the voting process, nine names have been replaced (see Figure 151). We observe that the roll mechanism for the selection of new high vol iBoxx sector index leads to more significant changes in the index as nearly one-third of the original entities in the high vol iBoxx index has been replaced. But we can conclude that the new selected entities should be a better representation of the higher volatility segment of the iBoxx composite index.

Figure 151. Deletions and Additions of the Entities in the High Vol iBoxx Index (CDX.NA.IG.HVOL), (Spreads in bp, Mid Levels), 18 Mar 04 Deletions Spread Replacements SpreadAmerican Electric Power Company, Inc. (AEP) 59 CenturyTel Inc. (CTL) 105AT&T Wireless Services, Inc. (AWE) 73 Citizens Communications Company (CZN) 330Capital One Bank (COF) 68 Comcast Cable Communications, Inc. (CMCSA) 81Computer Associates International, Inc. (CA) 134 Harrah's Operating Company, Inc. (HET) 81MBNA Corporation (MBNA) 64 Intelsat (INTEL) 85Sun Microsystems, Inc. (SUNW) 145 Kerr Mcgee Corp. (KMG) 81Toys "R" US, Inc. (TOY) 263 Pulte Homes, Inc. (PHM) 75Valero Energy Corporation (VLO) 62 Safeway Inc. (SWY) 73Visteon Corporation (VC) 251 Southwest Airlines Co. (LUV) 78

Average 124 Average 110

Source: Citigroup.



New Coupons

Dealers’ consortium also voted on the fixed rate coupons for each index. Based on the voting results, the new coupon for iBoxx CDX (CDX. NA.IG) has been set at 60bp per annum for five-year maturity (unchanged from March 2009 index coupon) and 80bp per annum for ten-year maturity (10bp wider than March 2009 index coupon). The new coupon for the high vol iBoxx CDX (CDX. NA.IG.HVOL) has been set at 115bp per annum for five-year maturity (10bp tighter than March 2014 index coupon) and 130bp per annum for ten-year maturity (5bp tighter than March 2014 index coupon).

Estimated Cost of the Roll and Market Implications

Based on the individual spread values of the deleted and replaced components of the iBoxx index (CDX.NA.IG), we can estimate the change in spread of the index. Six deleted names in the old iBoxx index have average spread equal to 160bp and six added names in the new iBoxx index have average spread equal to 73bp. After replacing the names, we estimate that the effect of change of entities should contribute to the spread adjustment for -4.2bp (this adjustment is based on the difference in average spread of replaced names scaled by the ratio of changed entities and total number of entities in the composite iBoxx index). In addition, as the new iBoxx index trades to maturity of September 20, 2009 (effectively with 5.5-year maturity), we estimate the curve adjustment for six months to be equal to 2.5bp. This curve adjustment has been estimated based on the spread difference between the five-year and ten-year iBoxx, taking into the account that the credit curve is steeper at the shorter end closer to five-year maturity. In total, we estimate that the spread adjustment for the roll of the iBoxx composite index (CDX.NA.IG) should equal to spread decrease of 1.7bp (see Figure 152).

Figure 152. Estimation of the Roll Effect on the iBoxx Index (CDX.NA.IG), (in bp, Mid Levels), 18 Mar 04 SpreadiBoxx CDX.NA.IG 3/09 62Adjustment for change of entitiesa -4.2Adjustment for the curveb 2.5Adjustment totalb -1.7iBoxx CDX.NA.IG 9/09b 60.3a Adjustment for change of entities is defined as difference between index spread contribution for deletions and replacement in the iBoxx index. b Estimate. Source: Citigroup.

We repeat the same estimation procedure of the roll effect also for the high vol iBoxx sector index. Nine deleted names in the old high-vol iBoxx index have average spread equal to 124bp and nine added names in the new iBoxx index have average spread equal to 110bp. After replacing the names, we estimate that the effect of change of entities should contribute to the spread adjustment for –4.3bp (again, this adjustment is based on the difference in average spread of replaced names scaled by the ratio of changed entities and total number of entities (9/30) in the high-vol iBoxx index). Adding the curve adjustments of 3bp (based on the estimation of the spread difference between the five-year and ten-year high-vol iBoxx index), we estimate that the spread adjustment for the roll of the iBoxx high-vol sector index (CDX.NA.IG.HVOL) should be equal to spread decrease for 1.3bp (see Figure 153).



Figure 153. Estimation of the Roll Effect on the High Vol iBoxx Sector Index (CDX.NA.IG.HVOL) (in bp, Mid Levels), 18 Mar 04 SpreadiBoxx CDX.IG.HVOL 3/09 116.5Adjustment for deleted entitiesa -4.3Adjustment for the curve b 3.0Adjustment total b -1.3iBoxx CDX.NA.IG.HVOL 9/09b 115.2a Adjustment for change of entities is defined as difference between index spread contribution for deletions and replacement in the iBoxx index. b Estimate. Source: Citigroup.

Comparison of March 2009 and September 2009 iBoxx Indexes In Figure 154, we represent the intrinsic spreads for the iBoxx index and each sector sub-index as based on the composition for the current March 2009 and new September 2009 maturity. The intrinsic value is defined as average spread of the individual spreads that comprise the reference portfolio. As we see in Figure 154, intrinsic for new iBoxx indexes trade at lower spreads, and that is a straightforward result based on the fact that the substituted names, (with the exception of AT&T Wireless) were those downgraded below the investment grade.

Figure 154. Intrinsic Spreadsa for Each of the IG iBoxx and Sector Indexes (in bp, Mid Levels), 17 Mar 04 Maturity

20 Mar 09 20 Sep 09 DifferenceCDX.NA.IG 69 65 -4CDX.NA.IG.CONS 63 57 -7CDX.NA.IG.ENRG 60 59 -1CDX.NA.IG.FIN 43 43 0CDX.NA.IG.INDU 73 67 -6CDX.NA.IG.TMT 99 95 -4CDX.NA.IG.HVOL 124 120 -4

a Intrinsic spread is defined as the average of the single-name CDS spreads included in the specific iBoxx index. Source: Citigroup.

We can observe the spread difference between the historical intrinsic value117 for the iBoxx CDX index with maturity March 2009 and the iBoxx CDX index with the maturity September 2009, (spread difference is defined as March 2009 maturity index spread minus September 2009 maturity index spread). As the goal of the roll process was to achieve a smooth transition, the historical intrinsics for the new iBoxx composite index (maturity September 2009) follows the intrinsics for the old iBoxx composite index (maturity March 2009) relatively closely (Figure 155, left panel). In the case of the high vol iBoxx index, we can observe that the intrinsic spread difference has been converging between two indexes, (Figure 155, right panel). That phenomena is driven by the fact that high volatility names in the (3/2009) iBoxx composite index have tightened over the course of the year. As the new high vol iBoxx was constructed based on the voting procedure, it should closely represent the currently more volatile names in the iBoxx index.

We also found the difference between the historical intrinsics for the old iBoxx sector indexes with maturity March 2009 and the historical intrinsics for the new iBoxx sector indexes with maturity September 2009 are in the range of 0bp–5bp over the course of the last year. That observation additionally indicates that the roll process for the iBoxx sector indexes should be relatively smooth and without major changes.

117 Historical intrinsic value is defined as the average of individual single-name spreads in the simulated iBoxx index as defined at its origination date.



Figure 155. Differences Between Historical Intrinsic Spreads for Simulated iBoxx Mar 09 and Sep 09 Indexes and Simulated High Vol iBoxx Mar 09 and Sep 09 Indexes, (in Basis Points, Mid Levels)

CDX.NA.IG Intrinsic Spread Difference CDX.NA.IG.HVOL Intrinsic Spread Difference

0

1

2

3

4

5

6

2 Apr 03 11 Jun 03 20 Aug 03 29 Oct 03 7 Jan 04 17 Mar 04-5

0

5

10

15

20

25

30

35

2 Apr 03 11 Jun 03 20 Aug 03 29 Oct 03 7 Jan 04 17 Mar 04

Source: Citigroup.

Note that the historical intrinsic spreads are not the same as the historical index spreads. Two factors tend to cause the intrinsic value to trade at a different (wider) level than the index. The first is that individual single-name CDSs trade with the Mod-R (modified restructuring) provision and the iBoxx indexes trade with No-R (no restructuring) clause. Second, the iBoxx indexes trade more frequently with a tighter bid-ask spread than the basket of underlying names. As a combination of these two effects, the market tends to trade the basis between the intrinsic value and the index spread between 3bp and 7bp.

Review of New Credits in iBoxx CDX.NA.IG

In Figure 156, we summarize the differences in credit quality for the six replacement credits to be added and those being removed from the IG iBoxx (the cohorts). We include a “Risk Rank” to each credit equal to the unweighted average of senior ratings from Moody’s, S&P, KMV, and Citigroup’s Value Opinion. For the Risk Rank, we assign a numerical score (AAA = 1; AA+ = 2; . . . BB- = 13) to each of the four ratings for each issuer and show each credit’s average Risk Rank along with the average for each cohort. The Citigroup Value Opinion is our Corporate Bond Research group’s assessment of credit risk while the Citigroup Outlook is their assessed credit trend over the next 12 months.

While we would expect the average credit quality of the replacement cohort to be higher (since the primary reason for replacement was credit specific ratings downgrades) — the magnitude of the improvement is noteworthy. Two single-A rated credits (ALLTEL and CVS) are added, replacing two credits with 5-B ratings or lower. The remaining four credits being added (Kerr McGee, Intelsat, CenturyTel, Supervalu) each have Risk Ranks comparable with the strongest credit in the replacement cohort, AT&T Wireless. As a result, the average Risk Rank for the six out of 125 entities that have been swapped improves notably from 10.8 to 8.4, which is equivalent to the difference between BB+ ratings and BBB+ ratings. On a whole IG iBoxx portfolio, this difference is barely noticeable.



Figure 156. Credit Profile for Swapped Issuers Included in IG iBoxx Roll REPLACEMENT CREDITS Senior Ratings

Citigroup Credit Underlying Credit Moody's S&P KMV Value Opiniona Risk Rankb Outlookc

ALLTEL Corp. A2 A A A- 6.3 Stable CVS Corp A2 A BBB A 6.8 Stable Century Telephone Baa2 BBB+ BBB- BBB 9.0 Stable Supervalu Baa3 BBB BBB+ BBB- 9.3 Stable INTELSAT Baa3 BBB+ NR BBB- 9.3 Developing Kerr McGee Baa3 BBB BBB- BBB- 9.8 Stable Average Risk Rank 8.4

REMOVED CREDITS Senior Ratings

Citigroup Credit Underlying Credit Moody's S&P KMV Value Opiniona Risk Rankb Outlooka AT&T Wireless Baa2 (Up) BBB (Dev) BB+ BBB+ 9.3 Stable Amerada Hess Ba1 BBB BBB BBB- 9.8 Negative Computer Associates Ba1 BBB+ (Dn) BB- BBB- 10.5 Stable Toys R Us Baa3 (Dn) BB (Dn) BB- BB+ 11.5 Stable Sun Microsystems Baa3 BB+ B BBB- 11.5 Stable Visteon Ba1 BB+ B BB 12.3 Developing Average Risk Rank 10.8 a Citigroup Value Opinion is Corporate Bond Research's assessment of credit risk and 12-month credit trend outlook. b Risk Rank is based on an evenly weighted average of Moody's, S&P, KMV, and Citigroup with AAA = 1; AA+ = 2; . . . BB- = 13).

Source: Citigroup.

APRIL 14, 2004

Glen Taksler Terry Benzschawel

Citigroup’s Primer on Credit Default Swaptions



Su

mm

ary

➤ Single-name credit default swaptions offer a way to express directional

views on credit or to hedge existing positions.

➤ A payer option is the right to buy CDS protection at a specified rate at some date in the future. Buying a payer is a bearish view on credit — investors make money if spreads widen.

➤ A receiver option is the right to sell CDS protection at a specified rate at some date in the future. Buying a receiver is a bullish view on credit — buyers of receivers make money if spreads tighten.

➤ We explain the difference between “knockout” and “no knockout” provisions in default, and explain how these provisions affect option payoffs.

➤ Straddles offer investors a way to express a view on CDS volatility.



A Primer on Single-Name Options

As trading activity in credit default swaptions increases, traditional credit market participants often find themselves confused by the meaning of “payer” and “receiver” options on CDSs. The purpose of this article is to provide an introduction to these products and show how they can be used to express directional views on credit.

Figure 157. Bullish and Bearish Views on Credit Using Options (Payoffs Assume No Default)

Bullish Bearish

Less Risk

Buy Receiver Own Call on Credit Make money if CDS narrows Maximum loss is premium

Buy Payer Own Put on Credit Make money if CDS widens Maximum loss is premium

More Risk

Sell Payer Short Put on Credit Keep premium if CDS narrows Maximum loss is notional

Sell Receiver Short Call on Credit Keep premium if CDS widens Maximum loss is price value of the strike — premium

See Figure 160 for payoffs in the event of default. Maximum loss on selling a payer option depends on whether there is a knockout provision (see Figure 160). The maximum loss on selling a payer would be less than the notional if (1) there were a knockout provision, or (2) there were a no-knockout provision, but positive recovery value. Source: Citigroup.

Payer Options Figure 157 shows four positions using options. A payer option is the right to buy CDS protection at a specified rate at some date in the future. If an investor buys a payer option, he makes money if CDS widens between now and expiry of the option contract.118 Buying a payer is therefore a bearish view on credit: If spreads widen (the credit deteriorates), the investor makes money. If spreads tighten (the credit improves), the most he can lose is the premium.

Typically, investors think of a payer as a put option on credit because as credit quality deteriorates the option becomes more valuable. Alternatively, a payer option can be viewed as a call option on spreads. An investor buys the right to purchase credit default protection at a pre-specified strike. As credit quality deteriorates, the CDS spread widens, and profit rises. One reason for using the term “payer” option as opposed to “put” or “call” is that a payer option is an option to pay a pre-specified rate for credit protection.

118 To be more precise, the investor makes money if, upon expiry of the option, the CDS exceeds the strike plus the premium paid for the option.

A payer option is the right to enter into a credit

default swap (CDS) at a specified rate at some

date in the future.

A payer option is, in effect, a put (on credit quality) and a call (on spreads).



Figure 158. Five-Year CDS Spread on Sprint, 28 Nov 03–7 Apr 04, and AT&T, 7 Jan 04–7 Apr 04 (In Basis Points) Sprint CDS AT&T CDS

80

90

100

110

120

130

140

150

28 Nov 24 Dec 19 Jan 14 Feb 11 Mar 6 Apr

Sprin

t CDS

(bp)

75

85

95

105

115

125

135

145

7 Jan 20 Jan 2 Feb 15 Feb 28 Feb 12 Mar 25 Mar 7 Apr

Source: Bloomberg.

Example — When to Buy a Payer

Suppose that an investor sold five-year credit default protection on Sprint (FON) in late November 2003, when it was trading around 140bp (see Figure 158, left). With Sprint trading at about 85bp (as of April 7, 2004), the profit would be 55bp. The investor believes that Sprint will continue to tighten but wants to hedge against the risk that credit spreads could widen. The investor buys a three-month at-the-money payer option. Thus, if Sprint widens, the losses from having sold the CDS will be offset by the gains from having bought the payer option, less the premium. If Sprint’s spread narrows, the investor also benefits, less the premium. Thus, buying a payer allows the investor to capture most of the upside but protects him from the downside risk.

Similarly, selling a payer is a bullish view on credit. Exposure to the underlying credit is the opposite of that in the case of buying a payer. The payoffs are opposite to those in the case of buying a payer. If the CDS spread tightens, the option expires out-of-the-money and the investor keeps the premium. By contrast, if the CDS spread widens, the investor loses money, and the maximum loss is only limited by the recovery value in default.

Example — When to Sell a Payer

Suppose an investor bought five-year credit default protection on AT&T (T) in early January 2004, when it was trading at about 80bp (see Figure 158, right). Currently, AT&T is trading at about 135bp (as of April 7, 2004), so the profit is 55bp. Assume also that the investor thinks that AT&T has room to widen a bit further, but is concerned that it might tighten. In this case, the investor should consider selling a slightly out-of-the-money payer option on AT&T, say at 140bp. If AT&T widens more than 5bp plus the spread value of the premium, the gains from having bought the CDS will be more than offset by having sold the payer option. The investor should be comfortable with locking in the upside, given his belief that AT&T spreads will widen only a bit further. If CDS spreads on AT&T tighten, the investor is protected so long as the premium received for selling the payer exceeds the losses on having bought the CDS. (If AT&T tightens significantly, he will still lose money.) In selling a payer option, the investor has extra cushion on the downside.

Selling a payer is a bullish view on credit.



Receiver Options Another way to express a bullish view on credit is to buy a receiver option. A receiver option is the right to sell CDS protection in the future at a given spread, in exchange for an upfront premium. A receiver option is, in effect, both a call (on credit quality) and a put (on spreads). If the CDS level on the underlying credit narrows, the value of the receiver option increases. A receiver option can also be viewed as a put on spreads because if credit quality improves, spreads narrow and the receiver option becomes more valuable.

Example — When to Buy a Receiver

Assume again that the investor buys protection on AT&T and made a profit as spreads widened. He thinks that AT&T might widen further but wants to protect his previous gains. Consider buying a three-month at-the-money receiver option. Should the CDS on AT&T widen, the profit would remain the same as that from the original CDS, minus the premium paid for the option. By contrast, should AT&T tighten, the losses on CDS would be offset by the increase in the value of the receiver, less the premium paid. Thus, going long a receiver option allows the investor continue to capture most of the upside, while protecting against downside risk.

If the investor sells a receiver, he is expressing a bearish view on credit. The payoffs are the exact opposite as when one buys a receiver. Should the CDS widen, the investor keeps the premium, because the buyer of the receiver option does not exercise. (If the investor were to exercise, he would receive a worse-than-market level for selling credit default protection.) However, should the CDS tighten, the seller of the receiver loses money, with the maximum loss being the price value of the strike, minus the premium. By price value of the strike, we are referring to the price change if the spread were to tighten to zero. This is what the strike of the option is worth when converted from spread (basis points) to dollars, or the strike times the forward DV01 of the credit.119 This brings us to the following point.

Effect of DV01 on Credit Swaption Payoffs

Although the payoff at expiry for an equity option depends only on the difference between the underlying asset (that is, the stock) price and the strike, credit swaption payoffs are dependent on the DV01 of the underlying CDS. This is because the option is struck in spread terms rather than on price, and the DV01 upon the expiry of the option (the “forward DV01”) converts from spread terms to dollar value. For example, suppose the investor buys an at-the-money June 2004 payer option for a premium of 40bp. If the DV01 of the five-year credit default swap in June 2004 is 4.5 ($4,500 per $10 million notional), then the investor makes money if the single-name CDS widens by more than 40bp divided by 4.5, or about 9bp. Thus, if the underlying spread were 70bp today, then the investor makes money if spreads widen to more than 79bp.

Alternatively, suppose that the investor were to sell an at-the-money June 2004 receiver option struck at 65bp for a premium of 15bp and the underlying credit default swap were to tighten all the way to zero. The amount that the investor would lose is 65bp times the 4.5 forward DV01 minus the 15bp premium, or about 278bp.

119 Practically speaking, the profit and loss change is limited by the maximum potential credit improvement.

A receiver option is the right to sell a credit default swap in the

future at a given spread, in exchange for

an upfront premium

If the investor sells a receiver, he is

expressing a bearish view on credit.

DVO1 is an important factor in the credit

swaption payoff.



Those familiar with options on corporate (cash) bonds may wonder why the DV01 conversion is not necessary on those options. This is because options on corporate bonds trade in dollar price, not spread, so no conversion is necessary.120

Example — When to Sell a Receiver

Consider again the Sprint example, in which the investor is short protection and spreads have rallied from 140bp to 85bp. The investor thinks that Sprint is unlikely to tighten more than 10bp and that there is only a small chance that it will widen back out. The premium from selling a slightly out-of-the-money receiver option struck at 75bp (the current 85bp, minus the 10bp you think Sprint has left to tighten) immediately adds to the profits. Should Sprint tighten more than 10bp, the receiver would be exercised, and the gains from having sold the CDS would be capped by having sold the receiver, but the investor would retain the swaption premium. On the other hand, should Sprint widen, the swaption premium would provide a cushion and further opportunity to liquidate your position without suffering losses.

Why should an investor consider selling an option (a more risky strategy), rather than buying one? First, because buying an option has a cost. In selling an option, the investor receives an upfront premium, at the cost of only partial downside protection. Provided the investor is attentive to the credit, hedging CDS by selling an option may provide sufficient cushion to unwind the position before it loses money.

Second, selling an option expresses a different view on potential spread movement than buying one. If the investor sells a receiver option to hedge a short CDS position, he has taken an implicit view that spreads will not widen dramatically. If they do, the investor can still suffer large losses. Similarly, a slight narrowing in spreads for an at-the-money option provides the same upside as a significant narrowing in spreads: the most the investor can earn is the option premium. So the upside is capped, while the downside is extensive. In short, investors who sell options should believe that spreads will trade in a narrower range than investors who buy options.

120 Corporate bond options are simply called calls and puts, and a call in dollar terms is the same as a call in terms of credit quality. A buyer of a call option on a corporate bond wants dollar price to go up and credit quality to improve. Similarly, a buyer of a put option on a corporate bond wants dollar price to go down, and credit quality to deteriorate. The seller has exactly the opposite payoff of the buyer.

Options on corporate (cash) bonds trade in

dollar price, not spread.

Selling a swaption is cheaper, but riskier

than buying one.



Figure 159. Payoffs From At-The-Money (182bp) Payer and Receiver Options on Five-Year Ford Motor Credit Corp (FMCC) CDS Based on a Notional of $10 Million, 7 Apr 04

Buy Receiver (Bullish) Put in Spread Terms, Call in Terms of Credit Quality

Buy Payer (Bearish) Call in Spread Terms, Put in Terms of Credit Quality

-200

-100

0

100

200

300

400

500

62 82 102 122 142 162 182 202 222 242 262 282 302

FMCC 5y Upon Expiry (bp)

Prof

it or

Los

s ($

Tho

usan

d)

-200

-100

0

100

200

300

400

500

62 82 102 122 142 162 182 202 222 242 262 282 302


Prof

it or

Los

s ($

Tho

usan

d)

Sell Receiver (Bearish) Put in Spread Terms, Call in Terms of Credit Quality

Sell Payer (Bullish) Call in Spread Terms, Put in Terms of Credit Quality

-500

-400

-300

-200

-100

0

100

200

62 82 102 122 142 162 182 202 222 242 262 282 302


Prof

it or

Los

s ($

Tho

usan

d)

-500

-400

-300

-200

-100

0

100

200

62 82 102 122 142 162 182 202 222 242 262 282 302


Prof

it or

Los

s ($

Tho

usan

d)

Source: Citigroup.

Figure 159 also presents payoff profiles versus spread levels for long and short payer and receiver swaptions on CDS for Ford Motor Credit Corp (FMCC). We assume a notional amount of $10 million and show the profit or loss in thousands of dollars. The horizontal lines show profits or losses for cases in which the swaptions expire out of the money. The diagonal lines show gains or losses for in-the-money spread levels. Notice that the payoff on the diagonal line is not one-for-one. This is because of the DV01 conversion from spread terms into dollar terms and the convexity of the price-yield relationship.

For the simplicity of an introductory piece, we do not go into a detailed discussion of convexity, since for small spread moves, convexity is not particularly important. Including convexity would reduce the payoff for significant spread widening on buying a payer option and increase the payoff for significant spread tightening on buying a receiver option.

Convexity exaggerates the payoff for significant

spread movements.



➤ If you were to buy a receiver option on FMCC, struck at 182bp, and spreads on option expiry (June 21, 2004) had narrowed to 82bp, the actual value of your position would be $360,347 per $10 million, rather than the DV01 approximation of $350,000, for a difference of $2,347. If you were to buy a payer option and spreads on option expiry had widened to 282bp, the actual value of your position would be $289,405, rather than the DV01 approximation of $325,000, for a difference of $35,595.

➤ The asymmetry (the smaller difference for 100bp of spread tightening than for 100bp of spread widening) is because the actual spread DV01 would be $4,700 per $10 million notional if spreads were to tighten 100bp — close to our approximation of $4,500. But this figure would be just $4,000 if spreads were to widen 100bp.

We emphasize that, for smaller spread moves, convexity is far less important.

Credit Swaption Payoffs in Default

Figure 160. Payoffs on Single-Name Options In the Event of Default

Bullish Bearish

Less Risk

Buy Receiver Lose premium

Buy Payer If no knockout, earn (100 – Recovery )% x notional — premium If knockout, lose premium

More Risk

Sell Payer If no knockout, lose (100 – Recovery )% x notional — premium If knockout, keep premium

Sell Receiver Keep premium

Source: Citigroup.

The payoffs described in Figure 157 are altered somewhat in the event of default as shown in Figure 160. That is, a single-name receiver option becomes worthless in default. The buyer of the receiver option loses the premium paid to the seller. The seller of the option keeps that premium. That is, it makes no sense for the buyer of the receiver option to exercise, as he would sell protection on a credit default swap on which he would owe par (due to default).

There are two possibilities for a single-name payer option in default. If the option carries a knockout in default provision, the option contract terminates out-of-the-money, with the buyer of the payer option losing the premium paid to the seller. However, if the payer option has no knockout provision, the option buyer will exercise, which entitles him to (100 — Recovery)% times the notional amount of the contract. That is, although the investor is out the option premium, in buying protection he receives par in exchange for delivering a cash bond. Because the payoff for the seller is exactly the opposite, the seller loses (100 — Recovery)% times the notional, but keeps the original premium.

In the event of default, the swaption payoffs become

more complicated.

If a payer option contains a “knockout in

default,” the contract becomes worthless.



➤ The difference in price between a swaption with a knockout provision and a swaption without a knockout provision is simply the upfront (or present value) cost of CDS until expiry of the option. In practice, knockout provisions have been more common, as they are not as sensitive to short-end CDS rates. Moreover, for short-dated options, investors are often comfortable with assuming that the underlying credit is unlikely to default within, say, the three-month term of the contract.

Credit Swaption Implied Volatility

The higher the volatility of a CDS spread, the greater the likelihood that an option written on that spread will finish in the money. Investors who have a view on CDS spread volatility, but no directional view, may wish to consider buying a straddle. Figure 161 shows payoff profiles for long and short straddles. A buyer of a straddle takes a view that underlying CDS volatility will be greater between now and expiry of the option contract than implied by its cost. In practice, if the CDS spread moves up or down by more than the premium divided by the forward DV01, then the buyer of an at-the-money straddle makes money. Conversely, a seller of a straddle believes that a firm’s CDS spread will be less volatile than the current implied volatility between now and expiry of the options contract.

Figure 161. Payoffs from Buying (Left) and Selling (Right) At-The-Money (182bp) Straddles on Ford Motor Credit Corp (FMCC) Based on a Notional of $10 Million, 7 Apr 04

-300

-200

-100

0

100

200

300

400

62 82 102 122 142 162 182 202 222 242 262 282 302


Prof

it or

Los

s ($

Tho

usan

d)

-400

-300

-200

-100

0

100

200

300

62 82 102 122 142 162 182 202 222 242 262 282 302


Prof

it or

Los

s ($

Tho

usan

d)

For the simplicity of an introductory piece, we ignore the effect of convexity. Source: Citigroup.

Conclusion We have presented an overview of payer and receiver options and provided examples of how investors can use these credit swaptions to express bullish and bearish views on credit and spread volatilities. We have also explained how the DV01 of the underlying CDS is important in calculating option payoff and demonstrated the effect of convexity on option value. Finally, we presented swaption payoffs in the event of default.

The value of the credit swaption contract is

closely tied to the implied volatility of its

underlying credit default swap.

Disclosure Appendix ANALYST CERTIFICATION For each security or issuer mentioned in this compendium report, the respective analyst (or analysts) who cover the security or issuer certifies that all of the views expressed in this research report accurately reflect the analyst's (or analysts') personal views about any and all of the subject securities or issuer(s). The analyst (or analysts) also certify that no part of the analyst's compensation was, is, or will be directly or indirectly related to the specific recommendation(s) or view(s) in this report.

Other Disclosures ADDITIONAL INFORMATION AVAILABLE UPON REQUEST Citibank, N.A., London Branch and Citigroup Global Markets Inc, including its parent, subsidiaries and/or affiliates (“the Firm”), may make a market in the securities discussed in this report and may sell to or buy from customers, as principal, securities recommended in this report. The Firm may have a position in securities or options of any issuer recommended in this report. An employee of the Firm may be a director of an issuer recommended in this report. The Firm may perform or solicit investment banking or other services from any issuer recommended in this report.

Within the past three years, the Firm may have acted as manager or co-manager of a public offering of the securities of any issuer recommended in this report. Securities recommended, offered, or sold by the Firm : (i) are not insured by the Federal Deposit Insurance Corporation; (ii) are not deposits or other obligations of any insured depository institution (including Citibank); and (iii) are subject to investment risks, including the possible loss of the principal amount invested. The Firm are regular issuers of, and trade in including position taking), traded financial instruments linked to securities which may have been reported on in the preceding research report.

Investing in non-U.S. securities, including ADR’s entails certain risks. The securities of non-U.S. issuers may not be registered with, nor be subject to the reporting requirements of, the U.S. Securities and Exchange Commission. There may be limited information available on foreign securities. Foreign companies are generally not subject to uniform audit and reporting standards, practices and requirements comparable to those in the U.S. Securities of some foreign companies may be less liquid and their prices more volatile than securities of comparable U.S. companies. In addition, exchange rate movements may have an adverse effect on the value of an investment in a foreign stock and its corresponding dividend payment for U.S. investors. Net dividends to ADR investors are estimated, using withholding tax rates conventions, deemed accurate, but investors are urged to consult their tax advisor for exact dividend computations.

Although information has been obtained from and is based upon sources the Firm believes to be reliable, we do not guarantee its accuracy and it may be incomplete or condensed. All opinions and estimates constitute the Firm 's judgement as of the date of the report and are subject to change without notice. This report is for informational purposes only and is not intended as an offer or solicitation for the purchase or sale of a security. This research report does not constitute an offer of securities. Any decision to purchase securities mentioned in this research must take into account existing public information on such security or any registered prospectus.

Investing in non-US securities by US persons may entail certain risks. Investors who have received this report from the Firm may be prohibited in certain US States from purchasing securities mentioned in this report from the Firm; please ask your Financial Consultant for additional details.This report is distributed in the United Kingdom by Citibank, N.A. London Branch or Citigroup Global Markets Limited, Citigroup Centre, Canada Square, Canary Wharf, London E14 5LB, UK. This material is directed exclusively at market professional and institutional investor customers and is not for distribution to private customers, as defined by the rules of the Financial Services Authority, who should not rely on this material. Moreover, any investment or service to which the material may relate will not be made available to such private customers. This material may relate to investments or services of a person outside of the United Kingdom or to other matters which are not regulated by the Financial Services Authority and further details as to where this may be the case are available upon request in respect of this material. If this publication is being made available in certain provinces of Canada by Citigroup Global Markets (Canada) Inc. ("The Firm Canada"), the Firm Canada has approved this publication. If this report was prepared by the Firm (excluding Nikko Citigroup Limited) and distributed in Japan by Nikko Citigroup Limited, it is being so distributed under license. This report is made available in Australia, to non-retail clients through Citigroup Global Markets Australia Pty Limited (ABN 64 003 114 832), and to retail clients through Smith Barney Citigroup Australia Pty Ltd (ABN 10 009 145 555), Licensed Securities Dealers. In New Zealand it is made available through Citigroup Global Markets New Zealand Limited, a member firm of the New Zealand Stock Exchange. This report does not take into account the investment objectives, financial situation or particular needs of any particular person. Investors should obtain advice based on their own individual circumstances before making an investment decision. Citigroup Global Markets (Pty) Limited is incorporated in the Republic of South Africa (company registration number 2000/025866/07) and its registered office is at Citibank Plaza, 145 West Street, Sandown, Sandton, 2196, Republic of South Africa. The investments and services contained herein are not available to private customers in South Africa. This publication is made available in Singapore through Citigroup Global Markets Singapore Holdings Pte Ltd, a licensed Dealer and Investment Advisor. This report is being distributed in Hong Kong by or on behalf of, and is attributable to Citigroup Global Markets Asia Limited, 20th Floor, Three Exchange Square, Hong Kong.

Citigroup Global Markets Inc. is a member of the Securities Investor Protection Corporation (SIPC). © Citigroup Global Markets Inc., 2004. All rights reserved Smith Barney is a division and service mark of Citigroup Global Markets Inc. and its affiliates and is used and registered throughout the world. Citigroup and the Umbrella Device are trademarks and service marks of Citicorp and its affiliates and are used and registered throughout the world. CitiFx® is a service mark of Citicorp. Any unauthorized use, duplication or disclosure is prohibited by law and may result in prosecution. Nikko is a service mark of Nikko Cordial Corporation.

FI04E078

the structured credit handbook - wiki-bazarjosh/tsb/creditderivitive... · collateralized debt...

Documents