SECOND EDITION
2006
Structured Credit Insights Instruments, Valuation and Strategies Primary Analysts:
Sivan Mahadevan Peter Polanskyj Vishwanath Tirupattur Pinar Onur Andrew Sheets
The Structured Credit Insights Series The Structured Credit Insights book, now in its second printing, contains 78 chapters on tranched credit instruments, markets, and investment strategies, many of which come from our Credit Derivatives Insights and CDO research publications. Section A, entitled "Getting Started," contains primers on both basic and state-of-the-art cash and synthetic CDOs, the tranched indices, first-to-default baskets, an intuitive guide to correlation, and common sensitivity measures. The remaining sections focus on valuation and sensitivity, investment strategies, market themes, performance, and cash and synthetic convergence themes. There is also a glossary for over 120 terms used in the market.
The Second Edition – What’s New? The second edition contains 34 new and numerous revised chapters focused on several topics. Recent default events and significant repricing of tranche relationships motivated chapters on valuation using both risk-neutral and realworld techniques. Continued growth in liquidity and breadth in the index tranche market resulted in the inclusion of material on investment strategies and relative value. A maturing market resulted in reports on new types of structures, including forward credit risk, managed portfolios and super senior tranches. Important regulatory capital and accounting policy shifts encouraged material on secular shifts in investment style. Synthetic innovation in the structured finance space prompted material on hybrid structured finance CDOs, and a booming leveraged loan market motivated the study of risks in the CLO market. We hope Morgan Stanley clients find this book useful, and we welcome any feedback so that we can improve future editions.
SECOND EDITION
2006
Structured Credit Insights Instruments, Valuation and Strategies
Primary Analysts: Sivan Mahadevan Peter Polanskyj Vishwanath Tirupattur Pinar Onur Andrew Sheets The Primary Analyst(s) identified above certify that the views expressed in this report accurately reflect his/her/their personal views about the subject securities/instruments/issuers, and no part of his/her/their compensation was, is or will be directly or indirectly related to the specific views or recommendations contained herein. This report has been prepared in accordance with our conflict management policy. The policy describes our organizational and administrative arrangements for the avoidance, management and disclosure of conflicts of interest. The policy is available at www.morganstanley.com/institutional/research. Please see additional important disclosures at the end of this report.
450
Morgan Stanley & Co. Incorporated April 21, 2006
Structured Credit Insights – Instruments, Valuation and Strategies
INTRODUCTION
1
SECTION A. GETTING STARTED: INSTRUMENTS AND PRIMERS 1
A CASH CDO PRIMER
10
2
A SYNTHETIC CDO PRIMER
26
3
A TRANCHED INDICES PRIMER
40
4
A FIRST-TO-DEFAULT BASKET PRIMER
48
5
WHAT IS CORRELATION?
56
6
UNDERSTANDING TRANCHE SENSITIVITY
66
SECTION B. VALUATION AND SENSITIVITY 7
A FRAMEWORK FOR SECONDARY MARKET CDO VALUATION
8
PORTFOLIO SAMPLING RISKS
112
9
STREETWISE CORRELATION
116
10
CORRELATION CONVERSATIONS, CONVEXITY IDEAS
120
11
HOW MANY SIDES IN A CDO SQUARED?
124
12
TAILORING BASKETS – ONE SIZE DOESN’T FIT ALL
128
13
LEARNING TO LIVE IN A SKEWED WORLD
132
14
CORRELATION – THE DELTA DAWN
138
15
TRANCHE STEPS – ONE BACK, TWO FORWARD
142
16
HIGH YIELD – THINKING OUTSIDE THE CORRELATION BOX
146
17
STRUCTURAL EVOLUTION – SURVIVAL OF THE FITTEST?
152
18
IS SEVEN CLOSER TO FIVE OR TEN?
158
19
DELPHI AND THE DELTA DIVIDE
164
20
READING INTO CORRELATION
170
21
TRANCHE OPTICS – THINGS SEEM BLURRY
174
22
REPEATING CORRELATION 101
178
23
CAN SPREADS OR CORRELATION GO TO ZERO?
184
82
SECTION C. INVESTMENT STRATEGIES 24
THE LONG AND SHORT OF IT
192
25
GETTING LONG SPREAD DECOMPRESSION
196
26
THROWING A CURVE AT CORRELATION
200
27
LESS RISK, BUT LONGER TAILS FOR TRANCHES
204
28
STREETWISE CORRELATION – TAKING THE HIGH ROAD
208
29
A SHOT AT HIGH YIELD TRANCHES – BOTTOM-UP
214
30
TRANCHES – NAVIGATING THE AUTO STORM
220
31
CORRELATION – DON'T THROW OPPORTUNITY OUT WITH THE MODEL
226
32
VALUING EQUITY TRANCHES IN THE REAL WORLD
236
33
HIGH YIELD EQUITY TRANCHES – ISN'T IT IRONIC?
242
34
90% FUNDAMENTALS – THE OTHER HALF IS CORRELATION
248
35
SUPER SENIOR SERENITY
254
36
THE GOOD STUFF IS ON THE TOP SHELF
260
37
SUPER SENIOR FRENZY
266
38
SIX MONTHS VS. SIX CREDITS
270
39
THE BIG PICTURE BEHIND RELATIVE VALUE
274
40
TRANCHE FORWARDS – THE NEXT LEG
280
41
SELF-MANAGED SYNTHETICS – SEEKING STABILITY
286
SECTION D. MARKET THEMES 42
SWINGING A DOUBLE-HEDGED SWORD
292
43
MEZZING WITH THE MARKETS
296
44
INSURANCE RESHAPES THE RISK PROFILE
300
45
CHALLENGING CORRELATION CYNICISM
304
46
BALANCING LIQUIDITY AND CREATIVITY
310
Structured Credit Insights – Instruments, Valuation and Strategies
SECTION D. MARKET THEMES (CONTINUED) 47
CORRELATION AND CREDIT CYCLES
316
48
HOW BIG IS THE STRUCTURED CREDIT MARKET? 2005 UPDATE
322
49
A LOAN IN THE STORM
328
50
LIEN ON ME, AGAIN
336
51
DAL + NWAC = CORRELATION WITHOUT A MODEL
342
52
WHAT DOES "BESPOKE" REALLY MEAN?
346
53
DELPHI – LET'S STEP OUTSIDE AND SETTLE THIS
352
54
SF CDOS: ABSOLUTELY SUBJECT TO CHANGE
358
55
ARE SENIOR TRANCHES STICKY?
372
56
TAKING A CLOSER LOOK
378
57
EUROPEAN CLOS: TOURING THE CONTINENT
394
58
DANA + DELPHI: WATCH THE COUNT AND THE ROLL
406
SECTION E. PERFORMANCE AND RATINGS 59
CDO PERFORMANCE – UNPLUGGED
414
60
LEVERAGE AND LONG-LASTING MILK
418
61
REVISITING CDO RATINGS MIGRATION
422
62
INTEREST RATES – RISING TIDES LIFT CDO BOATS?
430
63
CORRELATION OR CREDIT – WHAT DRIVES RATINGS CHANGES?
440
64
LOOKING BEHIND THE CURTAINS
446
65
TRANCHE PERFORMANCE – MEZZ SITS TIGHT & EQUITY TAKES A RIDE
452
66
MODIFICATIONS TO S&P SYNTHETIC CDO RATINGS MODEL
460
67
RATING BETWEEN THE LINES
464
68
CDO EQUITY PERFORMANCE – UNPLUGGED
470
SECTION F. CASH AND SYNTHETIC CONVERGENCE THEMES 69
BRIDGING THE HIGH YIELD GAP
480
70
A VIEW FROM THE TOP
484
71
ORIGAMI WITH CLO PAPER?
488
72
SEASONING FOR CONVERGENCE: THE FIRST STEPS
492
73
CASH & SYNTHETIC CDOS – SO WHERE'S THE CONVERGENCE?
502
SECTION G. FTD BASKET INVESTMENT THEMES 74
LEVERING THE LIGHTLY LEVERED
510
75
MANAGING MERGER MANIA
514
76
LEVERING THE FUNDAMENTALLY FIT
518
77
BASKET WEAVING 101
522
78
WHEN MERGERS DE-LEVER
526
SECTION H. GLOSSARY
532
Introduction Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Vishwanath Tirupattur Primary Analyst: Pinar Onur “We can’t solve problems by using the same kind of thinking we used when we created them.” – Albert Einstein As the structured credit markets enter their second decade of regular issuance and activity, we find that relatively recent events have served as important steps in the continued development of the market. There is clearly a secular trend in the application of both cash and synthetic structured credit in investment portfolios. For many investors, structured credit instruments represent solutions to challenges they face along many dimensions, including scale, liquidity, credit exposure limits, regulatory capital, liability matching, customization, globalization and, of course, human and technological resources. Market activity and new issuance over the past year have been considerable by any measure, with bespoke synthetic transactions over $600 billion annually and cash CDO issuance at over $200 billion. The term structured credit means different things to different people, but we tend to think of it quite literally. We define structured credit as the process of taking plain vanilla credit instruments and “structuring” them to meet certain goals, which can include diversification, loss or payment redistribution, hedging, principal protection, achieving ratings targets or stability, and altering a portfolio’s sensitivities to spread movements and defaults. Structured credit instruments can be focused on single names or portfolios, but for the purposes of this book, we think of structured credit in portfolio form and focused on corporate credit risk. We have addressed single names in a separate book (see Credit Derivatives Insights – Single Name Instruments and Strategies, 2nd Edition February 2006). Tranched credit, by far the largest portion of the structured credit business today, is really two markets that share a common lineage, one in which underlying collateral is cash assets (bonds and loans) and the other in which it is synthetic (credit default swaps). In our view, the two markets are conceptually more similar than they are different, but for many reasons that we shall discuss throughout this book, they have important structural differences and can be culturally disparate, as well. Nevertheless, we have devoted quite a bit of our research efforts over the years to both markets, and in this book we will discuss them from many different perspectives and also comment on how they might converge again in the future. WHAT’S IN A NAME?
Perhaps one of the most confusing aspects of the tranched credit market is the plethora of names used to describe and differentiate what is essentially a single idea. The French word tranche means slice, as in tranche de pain (slice of bread). If one thinks of the full loaf as being a blended credit portfolio, then the slices are the pieces that
Please see additional important disclosures at the end of this report.
1
Structured Credit Insights – Instruments, Valuation and Strategies
investors can use to gain exposure to the portfolio, although these slices can differ in terms of thickness, exposure to portfolio losses and cashflows.
Copyright 2004 Morgan Stanley
A commonly used product term that describes tranched credit risk is CDO, or collateralized debt obligation. As the term suggests, a CDO is a series of obligations that are dependent on the performance of an underlying portfolio (the collateral). There are many similarly used terms, including CBO, CLO, CSO and Synthetic CDO, which essentially differentiate bonds, loans, or synthetic collateral. As we address in great detail throughout this book, the correlation of default events is critical to valuing CDO products, and, as such, the term “correlation products” is often used synonymously with CDOs. Finally, while they look like credit default swaps on paper, first-to-default baskets (FTDs) certainly feel like tranched credit products from a risk perspective but differ from CDOs in important ways, which we will address in this book, as well. TRANCHES – THE CORPORATE CAPITAL STRUCTURE ANALOGY
Market participants tend to think of the various obligations of a CDO as a capital structure, much like the capital structure of a company. The bottom-most component is the equity, which has most of the upside of the collateral portfolio but is also the first to experience losses if the enterprise fails. As in common stock of a company, equity tranches of CDOs have limited liability, meaning that an investor’s loss is limited to the principal or notional exposure. The remaining obligations in a CDO’s capital structure are more “senior” than the equity, meaning that they have varying claims on the assets (remaining collateral) of the enterprise. These obligations can still be
2
relatively risky, the corporate analogies being everything from preferred stock to senior debt. The ultimate success or failure of a CDO is a function of the amount of “losses” that are incurred in the collateral portfolio, although the road to this ending point can be bumpy as well, given fluctuations in market prices of both the underlying collateral and tranches of the CDO. exhibit 1
The CDO Capital Structure
Corporate Capital Structure
Senior
Senior Debt
Subordinate Debt
CDO Capital Structure
Priority of Claim on Assets Rises with Seniority
Preferred Stock
Equity (Common)
Junior Senior Senior Mezzanine
Priority of Claim on Collateral Rises with Seniority
Junior Mezzanine First Loss or Equity
Source: Morgan Stanley
FINANCIAL ENGINEERING – PRIORITIZING PAYMENTS AND LOSSES
The simplest CDO structures are ones where payments to the various tranches are fixed, and the priority of losses follows a bottom-up pattern: the first losses go to the equity tranche, and then move up in the capital structures as losses grow beyond the “thickness” of the tranches. Today’s tranched index market (CDX tranches) follows this format, which makes their valuation relatively simple. But complexity increases, both in today’s state-of-the-art synthetic CDOs, and in cash CDOs, where the concept of a cashflow waterfall (top down) and triggers can make their “fair” valuation much more difficult. THE COMMON LINEAGE – A BRIEF HISTORICAL TOUR
The first CDO transactions date to the late 1980s, when Morgan Stanley and other investment banks literally bumped into each other at the steps of the ratings agencies in an attempt to securitize portfolios and distribute the resulting collateralized obligations in rated form. Marshal Salant, and the many who have worked with him at Morgan Stanley over the years, are generally credited for building the market for CDOs in the early 1990s, through structures that were referred to as repacks (repackaging of credit risk). Regular issuance of CDOs started in 1995, and there were many varieties. In the cash space, underlying collateral ranged from high yield bonds to middle market bank loans and sovereign emerging markets debt. Balance sheet CDOs referred to transactions where financial institutions moved bonds and loans from their balance sheets to the capital markets via securitizations. Arbitrage CDOs were structures where the difference in price between the assets of the portfolio and cost of the liabilities (of the CDO, including fees) left enough of a residual payment (allocated to the equity tranche) to make the exercise worthwhile. The “CDO arbitrage,” which is the excess risk premium in the market above and beyond some estimate of default risk, was (and, in many cases, continues to be) the motivation for the creation of CDOs. Moody’s introduced an important ratings model for CDOs in 1996.
Please see additional important disclosures at the end of this report.
3
4 Spread (bp)
Source: Morgan Stanley
Structured Credit Time Line
exhibit 2
0
200
400
600
800
1000
1200
Emerging Markets
High Yield
Investment Grade
First trust preferred CDO
FAS 133 becomes effective
Default correlation models gain popularity
Structured finance CDOs regular issuance begins
First arbitrage synthetic CDOs
Russian default causes turmoil in EM CDOs
Annual arbitrage CDO issuance tops $10 billion
2000 IG cashflow CBOs - regular issuance begins
1998 First IG cashflow CBO (Travelers Funding)
$280 billion notional credit risk referenced in synthetic structures
2002 the only down year for arbitrage CDO issuance
Regular issuance of IG CBOs ends in favor of synthetics
Regular issuance of HY CBOs ends in favor of CLOs
First hedge fund CFO
Largest managed synthetic CDO ($4.5 billion)
First managed synthetic CDO
HY default rates peak at 10.4% for this cycle
FASB addresses consolidation issues relating to SPEs
Delta-adjusted bespoke issuance $300 Bn
Basel II published - regulatory capital treatment for CDOs
Cash CDO issuance tops $100 Bn
First pure HY synthetic CDOs
Correlation model on Bloomberg introduced (CDSM)
Market adopts base correlation as a standard
Synthetic HY index tranches begin trading
Synthetic CDO-Squareds regular issuance begins
Fitch introduces correlation model
Industry standard Dow Jones CDX index introduced
SFAS 155 introduced, increasing US insurance co. and bank involvement
2006 Leveraged loan CDS standards emerge
Cash CDO issuance tops $250 Bn
Delta-adjusted bespoke issuance $600 Bn
Forward starting, self managed and CPPI structures emerge
S&P introduces significant changes to ratings model
High leveraged loan recoveries keep CLO ratings stable
Hybrid cash/synthetic ABS CDOs gain popularity
Delphi – 1st significant fallen angel default since 2002, in over 800 S&P rated synthetic CDOs
Delta Air Lines and Northwest file for bankruptcy within minutes of each other
Levered super senior products gain popularity
Collins & Aikman bankruptcy – 1st industry-wide settlement
2005 Auto sector stress spurs sell-off in index equity tranches
2004 Moody's adopts correlation models for rating synthetic CDOs
$950 billion notional credit referenced in synthetic structures
First structured credit hedge funds emerge
Synthetic IG index tranches begin trading
Synthetic TRACX Index introduced (100 names)
2003 Portfolio liquidation gives birth to an active cash CDO secondary market
2002 Synthetic Tracers launched
S&P adopts correlation models for rating synthetic CDOs
Popular press addresses defaults in HY CBOs
2001 First distressed debt CDO
HY Loans - regular use in CDOs
First balance sheet CLO
1996 Moody's introduces Binomial Expansion model for CDO ratings
First European HY CBO (EuroCredit)
First synthetic balance sheet CDO
First sovereign EM CDO Annual arbitrage CDO issuance tops $50 billion
1999 Synthetic balance sheet CDOs regular issuance
1997 EM CDOs - regular issuance begins
1995 HY CBOs - regular issuance begins
Structured Credit Insights – Instruments, Valuation and Strategies
In the cash CDO markets, 1997 marked the beginning of regular issuance of emerging markets CDOs, and the two years that followed saw quite a bit of emerging markets volatility. The short-lived history of investment grade cash CBOs started in 1998 and really ended in 2002 in favor of synthetic structures, while 1999 marked the first European cashflow CBO. In the same year, global issuance for arbitrage CDOs topped $50 billion, mainly due to CBOs backed by high yield bonds. The peak of high yield default rates for that credit cycle occurred in 2002 and led to the demise of high yield CBOs in favor of leveraged loan backed CLOs. This was the only year where cashflow CDO issuance was down on a year-over-year basis. The synthetic CDO market started in 1997 when the first synthetic balance sheet CDO was issued. Such structures were common during the late 1990s until many of the institutions (mainly banks) accomplished what they intended to do. At the same time, the underlying credit default swap market matured enough and investment grade spreads widened sufficiently to allow for the issuance of arbitrage synthetic CDOs in 2000. A key year for the synthetic CDO market, 2002 witnessed numerous fallen angel defaults causing pricing stress, leading to the popularity of managed structures. The structured credit market, in its various forms, was largely a bespoke market, meaning that each transaction was unique, but the market was fragmented and investors never expected much secondary market liquidity. Yet in early 2003, two events served to inject a fair amount of liquidity into the secondary market. In the cash CDO markets, a financial institution with a large portfolio of senior notes hired a money manager to help liquidate its CDO portfolio, an action which overnight brought in many niche buyers and jump-started a secondary market for cash CDOs. At the same time, in the synthetic space, an even more dramatic event occurred. The growth of the credit default swap market permitted the creation of a 100-name standard index (Dow Jones TRAC-XSM) traded by multiple dealers, the first time such a broad index had been created. Dealers began making markets in standardized tranches of this index, providing for the first time a two-way market and important pricing transparency. The term “correlation,” previously a concept discussed only by analytics providers and risk managers, began the process of becoming more mainstream. The index tranche market itself has evolved quite a bit, with today’s products centered around the Dow Jones CDX family of indices. With tranched indices in place, nearly $1 trillion of notional credit risk was referenced through synthetic structures in 2003. In 2004, volumes almost doubled, with $1.8 trillion of credit risk referenced through synthetic structures, while structural complexity increased dramatically. The rating agencies became more aggressive about catching up to a market that passed them by. Synthetic high yield CDOs emerged in 2004, as well. The early part of 2005 saw a continuation of these themes, particularly on the structural evolution front. 2005 proved an important crossroads for the synthetic structured credit market. Stress in the US auto sector served as a big test for the market, resulting in significant enough equity tranche volatility to grab the financial media spotlight briefly. The structured credit market recovered, and numerous credit events played important operational roles. The bankruptcy filing by Collins & Aikman (a US auto supplier) resulted in the first industry-wide settlement of index tranches using an ISDA coordinated global auction. The Delphi default was significant given that Delphi appeared in over 800 bespoke structures, representing the first US fallen angel default since 2002. In a unique
Please see additional important disclosures at the end of this report.
5
Structured Credit Insights – Instruments, Valuation and Strategies
example of real world default correlation, airline operators Delta and Northwest filed for bankruptcy within minutes of one another. In the early part of 2006, two significant market themes emerged: the implementation of Basel II, which supports taking credit risk in tranched or securitized form; and FAS 155, which potentially treats credit linked notes containing synthetic structured credit similar to cash corporate bonds from an accounting perspective. Recent themes in the cash CDO market include the emergence of hybrid cash/synthetic structured finance CDO structures, including those that are fully synthetic, given standardization of CDS linked to structured finance instruments. In the CLO space, continued high recoveries have surprised many participants, and the development of leveraged loan CDS is encouraging fully synthetic CLOs, as well. THE TRANCHED CREDIT MARKET TODAY
Today, the index tranche market is very much the “on-the-run” market for synthetic CDOs. Many billions of dollars in notional risk are traded each week through benchmark products in the US, Europe, and Asia, spanning investment grade and high yield markets. The index tranche market has introduced valuation transparency to a point where a standardized correlation model has emerged. Market implied correlation has become an important basis for valuation, but alternative techniques are quite popular as well. Liquidity in off-the-run index tranches has broadened the index tranche market quite substantially, resulting in an interesting pool of relative value ideas. As liquidity and transparency developed in the index tranche market, it benefited the bespoke market as well, providing a way for arrangers to hedge market risk and motivating the creation of perhaps more complicated structures that are better suited to meet the risk and return goals of specific investors. Structures where underlying collateral are CDO tranches themselves (referred to as CDO-squared transactions) are common today, as are structures that offer some form of principal protection, fix recoveries, and allow for subordination levels to adjust, as well as those linked to other markets (interest rates, equities, and commodities) in different ways. In the corporate credit space, the cash CDO market has been dominated by structures backed by leveraged loans. These CLO structures have encouraged the leveraged loan market itself to grow. Shorter maturities, more debt seniority, and higher historical recoveries have made CLOs far more accepted in the market place versus the high yield bond backed CBOs that dominated the cash CDO markets several years ago. While correlation models are widely used in the synthetic structured credit markets today, we have witnessed a market that has tested the “risk-neutral” correlation models, with many standard tranches trading at quite a bit of “skew” to the traditional models, driven by falling risk premium and the importance of credit ratings. Some of this skew can be explained by risk premium in the market and the importance of secular shifts based on ratings and accounting policy. There are also differences in correlation model usage, which is another source of skew. We have devoted quite a bit of our research to the balance of risk-neutral correlation models with the secular shifts in investment styles.
6
WHERE IS THE MARKET GOING?
The amount of credit risk in structured credit instruments today is quite large, as is the investor base. In the structured credit markets, we see some parallels to the concern and cynicism regarding single-name default swaps years ago. Structured credit instruments have brought in (or perhaps even created) an investor base that is more focused on capturing and even leveraging the risk premium in the market over and above some measure of default risk. This process, we argue, is a secular change for the corporate credit markets but has been in the making for years. The net result is more leverage in the credit market and more investor diversity; resulting volatility of the credit market has increased, but perhaps so has the liquidity. Analogies to the mortgage-backed securities markets and even the interest rate and equity markets can easily be drawn. THE STRUCTURED CREDIT INSIGHTS SERIES
The Structured Credit Insights book, now in its second printing, contains 78 chapters on tranched credit instruments, markets, and investment strategies, many of which come from our Credit Derivatives Insights and CDO research publications. Section A, entitled “Getting Started,” contains primers on both basic and state-of-the-art cash and synthetic CDOs, the tranched indices, first-to-default baskets, an intuitive guide to correlation, and common sensitivity measures. The remaining sections focus on valuation and sensitivity, investment strategies, market themes, performance, and cash and synthetic convergence themes. There is also a glossary for over 120 terms used in the market. THE SECOND EDITION – WHAT'S NEW?
The second edition contains 34 new and numerous revised chapters focused on several topics. Recent default events and significant repricing of tranche relationships motivated chapters on valuation using both risk-neutral and real-world techniques. Continued growth in liquidity and breadth in the index tranche market resulted in the inclusion of material on investment strategies and relative value. A maturing market resulted in reports on new types of structures, including forward credit risk, managed portfolios and super senior tranches. Important regulatory capital and accounting policy shifts encouraged material on secular shifts in investment style. Synthetic innovation in the structured finance space prompted material on hybrid structured finance CDOs, and a booming leveraged loan market motivated the study of risks in the CLO market. We hope Morgan Stanley clients find this book useful, and we welcome any feedback so that we can improve future editions. ACKNOWLEDGEMENTS
We would like to acknowledge the many other Morgan Stanley strategists who were involved in some of the research ideas in the various chapters of this book, including Nikhil Agarwal, Anisha Ambardar, Brian Arsenault, Lindsey Burnett, Payal Dave, Viktor Hjort, Rizwan Hussain, Young-Sup Lee, Rachel Ng, Greg Peters, Simmi Sareen, Gayatri Talreja, and Sejal Vora. We would also like to acknowledge our structured credit businesses within Morgan Stanley whose team members have played a significant role in making the tranched credit markets more liquid and transparent over the years, permitting investors and researchers to take a strategic approach to this market. We would also like to thank our creative services team, especially Dan Long, Patrick Oliver and Linda Warnasch, for their assistance in preparing this book. Finally, we very much appreciate the immeasurable amount of feedback that we have received on our reports from Morgan Stanley clients.
Please see additional important disclosures at the end of this report.
7
Section A
Getting Started: Instruments and Primers
Please see additional important disclosures at the end of this report.
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 1
A Cash CDO Primer INTRODUCTION
The emergence of a market for Collateralized Debt Obligations (CDOs) is a significant development in the evolving area of securitization of credit risk. The cash CDO 1 market has now been in existence for more than 15 years, with consistently substantial issuance volumes since the mid 1990s and is firmly established as a mature asset class having withstood the peaks and troughs of a full credit cycle (Exhibit 1). As depicted in Exhibit 2, CDOs have emerged as a major segment of the overall asset-backed securities market as well. exhibit 1
Growth in CDO Issuance
$ Billion
Transaction Volume ($Bn)
300
Number of Transactions
Number of Transactions 600 268
250
500
200
400
150
300 135
100 83 50 0
91 80
200
99
91 73
78
49 20
100 0
1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 (YTD)
Note: YTD 2006 data is up until March 2006. Sources: Thomson Financial Service, Moody’s Investors Service, Fitch IBCA, Inc., Asset-Backed Alert, MCM Corporate Watch, S&P, Morgan Stanley
While the CDO market started as an efficient mechanism for managing credit risk on bank balance sheets and for obtaining regulatory capital relief, CDOs have evolved into complex instruments to achieve leveraged returns to investors with a wide range of credit risk appetites. Today’s CDOs encompass a vast array of underlying assets ranging from unsecured debt instruments such as high grade, high yield and emerging market bonds; secured debt instruments such as middle market and leveraged loans; subordinated instruments such as trust preferred securities, to all forms of structured finance obligations including ABS, CMBS and RMBS, hedge fund obligations and finally CDO tranches themselves. Exhibit 3 illustrates the diversity of CDO collateral by providing a comparison of the collateral composition of new issue CDOs between 1998 and 2005.
1
While CDO is an all encompassing term, it is common to use the terms Collateralized Bond Obligation (CBO), Collateralized Loan Obligation (CLO) and Collateralized Fund Obligation (CFO), depending upon the predominant assets in the collateral portfolio – bonds, loans and hedge funds, respectively.
10
Please see additional important disclosures at the end of this report.
11
0
100
200
300
400
500
600
CMBS
Credit Card ABS
CDOs Home Equity Loan ABS
Auto Loan ABS
1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002
Volume ($Bn)
Note: (1) As of December 31, 2005. All issuance numbers are global. Sources: Moody’s Investors Service, Fitch IBCA, MCM Corporate Watch, Asset-Backed Alert, Morgan Stanley
CDOs in the Context of Overall ABS Market
exhibit 2
2003 2004 2005
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 1
1998
exhibit 3
A Comparison of Collateral Composition of CDO Issuance – 1998 and 2005
Investment Grade 2.5%
High Yield Bonds 46.5%
Structured Finance 7.3% Emerging Markets 6.2%
High Yield & Middle Market Loans 35.8%
Other 1.7%
2005 Trust Preferred 3%
Structured Finance 44%
Other 8% EM 1%
Middle Market Loans 18%
HY Loans 26%
Source: Thomson Financial, Moody’s Investors Service, Standard & Poor’s, Fitch/IBCA, MCM Corporate Watch, IFR
While CDO issuance using credit derivatives technology has become popular over the last few years (see Chapter 2 for a primer on synthetic CDOs), the issuance of CDOs in their original form as cash instruments remains substantial. In this chapter, we focus on cash CDOs to provide an overview of the basics of types, structures, motivations and performance metrics of cash CDOs.
12
WHAT ARE CASH CDOs?
CDOs are stand alone special purpose vehicles (SPV) that invest in a diversified pool of assets. These investments are funded through the issuance of multiple classes of securities, the repayment of which is a function of the performance of the pool of assets which serve as collateral. These classes of securities are sold to investors who assume the risks of the pool of assets in different orders of seniority, with each class representing a priority of cash flows generated by the underlying collateral. When the SPV purchases the pool of assets outright, as opposed to acquiring risk exposure using credit derivatives, the CDOs are called cash CDOs. HOW DO CASH CDOs WORK?
Exhibit 4 provides a diagrammatic representation of the working of a cash CDO. In this simplified illustration, the SPV issues three classes of securities, all at par – one class each of senior notes, junior notes and preferred shares. 2 The senior and junior notes are the debt tranches and the preferred shares constitute the equity tranche of the transaction. The proceeds are used to purchase a diversified collateral pool of assets. Cash flows from the assets are used to pay the manager and trustees of the transaction and make principal and interest payments to the note holders in the order of seniority – senior notes first, followed by the junior notes. Preferred shares are the residual of the transaction and receive a current coupon out of the residual interest proceeds generated by the collateral. The coupon may be deferred or eliminated depending upon available cash flow. As assets in the portfolio default, residual cash flows and consequently, the payments to the preferred shareholders decrease. If defaults reach a certain level, the principal amounts invested by the preferred shares may be written down, followed by the principal of the junior notes and so on. The coupons on the senior notes are set to be lower than the coupons on the junior notes reflecting the lower risk assumed by the senior note holders. The preferred shares represent a leveraged investment in the underlying collateral pool of assets. Being in a first loss position, they have the highest risk exposure and consequently the highest return potential. They also face higher volatility of return than the underlying collateral pool. In some cash CDOs, the collateral pool of assets may predominantly have fixed rate coupons while the liabilities may be predominantly floating rate. This mismatch introduces interest rate risk into the transaction, which is usually addressed through an interest rate hedge (fixed-floating rate swap or an interest rate cap/swaption) at deal inception. The notional amount of the hedge is a function of the expected amortization of the floating rate liability tranches. If the realized amortization differs from the expected amortization at deal inception, the hedge itself may introduce a degree of risk into the transaction. We explore this topic in detail in Chapter 62.
2
Throughout this chapter, we use the terms preferred shares and equity tranche interchangeably.
Please see additional important disclosures at the end of this report.
13
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 1
In some transactions, the SPV enters into an insurance contract with a monoline bond insurance company to purchase external credit enhancement. The bond insurer guarantees the payment of principal and interest on one or more classes of the debt tranches. The notes so insured are said to have been “wrapped.” Usually, the ratings of these notes reflect the financial strength rating of the bond insurer. The assets are purchased and managed by a collateral manager in accordance with a set of guidelines, which are designed to provide investors with an exposure to a diversified pool of assets. Typically, trading within the portfolios ceases 3-5 years after the CDO issuance. Around that time, the debt tranches start to amortize. Structural mechanisms are put in place to protect the integrity of the capital structure and to maintain the balance between the interests of the debt and equity tranches. We discuss the structural mechanisms in detail in a later section of this chapter. exhibit 4
How a Typical Cash CDO Works
External Credit Enhancement (Bond Insurer)
Interest Rate Hedge
Aggregate Principal & Interest
Diversified Collateral Pool of Assets Proceeds
CDO Special Purpose Vehicle (SPV)
Principal & Interest
Principal & Interest Proceeds Residual Cash flow Proceeds
Trustee
Senior Notes
Proceeds
Junior Notes Preferred Shares
Collateral Manager
Source: Morgan Stanley
Each transaction also involves a trustee who acts as the custodian responsible for the safe custody of the assets and for ensuring compliance with the trading guidelines and other structural features. Frequently, the trustee is also the calculation agent for the transaction responsible for computing the payments due to the different parties in the transaction according to deal documentation. Periodically, the trustee provides reports to investors regarding the status of the CDO’s assets and liabilities as well as compliance with regard to structural mechanisms specified in the CDO documentation at deal inception. These reports are critical for the ongoing monitoring of a transaction. Finally, debt tranches are typically callable at the option of the preferred share holders after a stated non-call period, usually 2-5 years. The call option embedded in senior CDO notes is generally a Bermuda-style option, in that the call is exercisable on discrete exercise dates after the non-call period. Structurally, the embedded callability is intended to provide a means for equity investors to unwind the CDO transaction
14
once it has de-levered to the point that on a forward-looking basis, returns are no longer deemed attractive. In addition, the call enables equity investors to realize capital appreciation in the underlying collateral pool. In some cases, callability involves makewhole premiums, which require payments greater than the par amount of outstanding liabilities of the CDO. CASH CDOs: A TAXONOMY
CDO investors are faced with an overabundance of terminology to describe different types of structures. In this section, we provide simple explanations of some key terms. Depending upon the motivation behind the transaction, cash CDOs can be categorized into two types – balance sheet and arbitrage transactions. Balance sheet transactions are intended to obtain regulatory and/or economic capital relief for financial institutions holding bonds and/or loans, and achieve a higher return on assets through redeployment of capital. The financial institution often retains the equity tranche of the CDO. Usually, the assets securitized through balance sheet CDOs are already on the balance sheets of these financial institutions; therefore, balance sheet CDOs require very short ramp-up periods. In addition, during the life of the transaction, they experience only limited trading in the underlying collateral pool of assets. Arbitrage transactions are motivated by the aspirations of the equity tranche investors and the collateral managers. The former seek to achieve a leveraged return as the spread between the post-default yield of the collateral pool of assets and the cost of financing the assets through the issue of the debt tranches. This spread, also known as the funding gap, is the arbitrage that the equity investors are seeking to capture. The collateral managers seek to expand assets under management in order to realize fees for the management (acquisition, trading and monitoring) of the collateral pool of assets. 3 Depending upon the mechanics of structural protection to the debt tranches, cash CDOs are categorized into cash flow and market value structures. Cash-flow structures are based on the ability of the collateral to generate sufficient cash flow to pay the coupon and principal on the debt tranches. If the credit quality of the collateral decreases below certain specified levels, cash flows are diverted from subordinated tranches to the senior tranches to pay their coupons. If certain triggers are breached, cash flows are diverted to repay the principal of the senior tranches in an accelerated fashion until the metrics of collateralization revert to the levels above the triggers. Market value structures, in contrast, depend upon the ability of the fund manager to maintain the market value of the collateral to pay the CDO debt tranches. In a market value transaction, assets in a CDO’s collateral pool are valued (marked-to-market) periodically, incorporating cushions to account for future variability in the market value of the assets. If the value of the assets falls below the sum of the par amounts of the debt tranches, some assets are sold and a part of the debt tranches is repaid until the market value of the assets exceeds the par value of the remaining debt tranches.
3
In order to ensure an alignment of interests, collateral managers often invest in a portion of the equity tranches and subordinate a portion of the management fees to the debt and equity tranches.
Please see additional important disclosures at the end of this report.
15
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 1
Though it was not always the case, currently a vast majority of cash CDOs are arbitrage transactions. Exhibit 5 contrasts the composition of the new issue CDO market composition in 1998 and 2005. 1998
exhibit 5
Composition of New Issue CDO Market – 1998 and 2005
Balance Sheet 48%
Arbitrage 52%
2005
Balance Sheet 18%
Arbitrage 82%
Sources: Thomson Financial Service, Moody’s Investors Service, Fitch IBCA, Inc., Asset-Backed Alert, MCM Corporate Watch, S&P, Morgan Stanley
While both cash flow and market value structures are used in arbitrage CDOs, for the most part, balance sheet CDOs deploy cash flow structures. Even within arbitrage CDOs, cash flow structures constitute the vast majority of the currently outstanding transactions. A key determinant of a CDO structure pertains to the amount of leverage within a structure. Leverage refers to the size of the equity tranche relative to the total size of the transaction. For example, if the equity tranche is $100 million in a CDO of $1 billion, it is said to have a leverage equal to 10 times. While there is a significant variation in the degree of leverage in different transactions based on the nature of risk in the collateral portfolio, arbitrage cash flow transactions generally have a higher degree of leverage than market value transactions (8-12 times versus 4-6 times) and balance-sheet cash-flow transactions have much higher levels of leverage (25-50 times).
16
AN OVERVIEW OF STRUCTURAL FEATURES AND PERFORMANCE TESTS IN CASH CDOs
In this section, we focus on arbitrage cash flow CDOs to describe typical structural features and performance tests that determine the priority of payments 4 designed to maintain the integrity of CDO capital structure. Senior notes in a cash flow CDO transaction have a priority claim on all cash flows generated on the underlying collateral pool and are protected by subordination, overcollateralization and coverage tests, which serve to accelerate the redemption of the senior notes if the tests are violated. A number of other structural features and limitations are imposed to ensure that the collateral pool is well diversified and remains consistent with the intended credit quality at deal inception. Subordination: The priority of claims of the senior notes means that the claims of junior notes and the equity tranches are subordinated, in that order. The size of the junior and equity tranches therefore describes the amount of subordination available to the senior note holders. The amount of subordination is a function of the credit quality of the underlying pool of assets, their expected losses given default and the desired rating. The lower the credit quality of the underlying pool of assets, the higher the level of subordination required for the senior notes to obtain a desired rating. Overcollateralization (O/C) and Coverage Tests: The mechanisms most used to ensure the priority of payments are various collateral coverage tests. Overcollateralization is the excess of the par amount of the collateral available to secure a class of CDO notes. If $100 of par assets are available to service $80 of senior notes, the ratio 100/80 = 125% suggests that the senior notes are overcollateralized by 25%. Generically, senior and junior par coverage tests 5 are calculated as shown below. Senior Tranche Par Coverage Test =
(Par Value of Performing Assets + Adjusted Par Value of Defaulted Assets 6) / Par Amount of the Senior Notes
Junior Tranche Par Coverage Test =
(Par Value of Performing Assets + Adjusted Par Value of Defaulted Assets) / Par Amount of the Senior and Junior Notes
4
The priority of payments is also referred to as the cash flow waterfall. The par coverage tests are also called O/C tests. 6 Adjusted par value of defaulted assets reflects expected recoveries for defaulted assets whose final valuation is yet to be determined. 5
Please see additional important disclosures at the end of this report.
17
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 1
Along the same lines, interest coverage tests are calculated as: Senior Tranche Interest Coverage Test =
Interest expected to be collected on the performing assets for the current period / Interest due on the senior notes for the current period
Junior Tranche Interest Coverage Test =
Interest expected to be collected on the performing assets for the current period / Interest due on the senior and junior notes for the current period
Trigger levels for each of these tests are covenanted in the deal documentation. In general, junior notes will have par coverage triggers set at lower levels than the senior notes. At deal inception, the actual values for each of these tests will be well above the trigger levels. A breach of the coverage tests below the trigger levels results in additional trading restrictions and/or potential diversion of cash flows as identified in the deal indenture’s priority of payments (see Exhibits 6 and 7). If the par coverage test for the senior tranche is breached, the senior notes will be redeemed until the test comes back into compliance. Likewise, if interest coverage tests are breached, all interest payments will be redirected to the senior bonds until the trigger is cured. In general, interest coverage tests are relatively less onerous than par coverage tests. Still, the presence of these tests contributes to maintenance of higher levels of current coupon income than would otherwise be the case. The levels at which O/C test triggers are set are an important consideration in a cash CDO, given their effect on the cash flow waterfall. For each class of notes, they should be set at levels high enough to ensure that the class can withstand a certain level of defaults in the underlying pool of assets. The higher the trigger level, the more likely it is that a given level of default will cause the O/C trigger to be breached. From the perspective of investors of subordinated tranches, O/C triggers set too high increase the likelihood that cash flows are diverted away from them. Ideally, transactions should be structured such that O/C tests for the junior notes are triggered before the senior notes. Another structural provision frequently seen in CDOs is the payment-in-kind provision (“PIKing”). Some CDOs provide for the deferment of interest to subordinated debt tranches upon the breach of certain coverage tests. PIKing results in the deferred interest being added to the principal of those tranches, where it earns interest at the same rate as the original principal. Supplemental O/C Test: During the last downturn in credit cycle (2001-02), several CDO notes experienced downgrades that were attributed to their structural weaknesses. The argument made was that the trigger levels of par coverage tests were set at levels so low that a violation would occur only after significant deal deterioration. Such criticism was addressed by the introduction of a Supplemental O/C test that becomes binding before the par coverage tests are breached by setting its trigger level higher than the par coverage test triggers. A violation of Supplemental O/C tests results in
18
reinvestment of excess cash flow rather than redemption of the senior notes. Some part of collateral manager fees and payments to the equity tranche are not paid until the test is cured. This mechanism enables a collateral manager to build up overcollateralization over time and offset downward pressure on ratings without paying down the least costly portion of a CDO’s capital structure. Excess CCC Haircut: Another issue that became prominent during 2001-02 was the extent of CCC-rated assets in CDO portfolios because of severe downgrades in the corporate sector. These securities were still performing but were highly likely to experience potential par losses in the near term as their downgraded ratings suggested. Nevertheless, for the purpose of the computation of par coverage tests, they were being given full par credit. In some cases, collateral managers were resorting to “barbelling” of the credit portfolio – buying deeply discounted CCC-rated bonds whose risk potential was not being reflected in the coverage tests. The excess CCC haircut was a proviso introduced to address this issue. In short, CCC-rated securities in excess of some predetermined level would be treated at their market value instead of at par value. exhibit 6
Interest Waterfall of a Sample CDO
Colla te Poo ral l
Trustee and Administrative Fees
Senior Management Fee
If Coverage Tests Are Met
Interest on Senior Notes
Interest on Mezzanine Securities
Subordinated Management Fee
Residual to Subordinated Notes
If Coverage Tests Are Not Met (1) Redemption of Senior Notes
After Senior Notes Have Been Fully Redeemed
Redemption of Mezzanine Securities
After Mezzanine Securities Have Been Fully Redeemed
Residual to Subordinated Notes
Note: (1) If coverage tests are not met, and to the extent not corrected with principal proceeds, the remaining interest proceeds will be used to redeem the most senior notes to bring the structure back into compliance with the coverage tests. Interest on the mezzanine securities may be deferred and compounded if cash flow is not available to pay current interest due. Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
19
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 1
exhibit 7
Principal Waterfall of a Sample CDO
Coll ate Poo ral l
Trustee and Administrative Fees (1)
Senior Management Fee (1)
Interest on Senior Notes (1)
Interest on Mezzanine Securities (2)
During Reinvestment Period and for Unscheduled Principal after Reinvestment Period.
Reinvestment
For Scheduled Principal Payments after Reinvestment Period or if Coverage Tests are not met.
Redemption of Senior Notes
Redemption of Mezzanine Securities
Subordinated Management Fee
Residual to Subordinated Notes
Note: (1) To the extent not paid by interest proceeds. (2) To the extent senior note coverage tests are met and to the extent not already paid by interest proceeds. If coverage tests are not met, the remaining principal proceeds will be used to redeem the most senior notes to bring the structure back into compliance with the coverage tests. Interest on the mezzanine securities may be deferred and compounded if cash flow is not available to pay current interest due. Source: Morgan Stanley
COLLATERAL QUALITY TESTS
Unlike the coverage tests, the purpose of quality tests in CDOs is not to redirect cash flows in the waterfall but to ensure that the composition of the portfolio does not change drastically over time. When the quality tests are breached, trading within the portfolio becomes restricted. Some of the frequently used collateral quality parameters are minimum diversity score, maximum weighted average rating factor (WARF), issuer concentration limits, maturity and weighted average life (WAL) limits and weighted average spread (WAS). Diversity Score is a statistic developed by Moody’s 7 to reflect the degree of diversification within the collateral pool. Assets in the collateral pool are mapped into a hypothetical portfolio of N number of uncorrelated, homogeneous assets with identical default probabilities and equal par values. The number N in the hypothetical portfolio is the “diversity score.” The higher the diversity score, the more diversified the reference portfolio. The formula for calculating the diversity score incorporates obligor
7
See “The Binomial Expansion Method Applied to CBO/CLO Analysis,” Moody’s Investor Service, December 13, 1996 and “Moody’s Approach to Rating Multi Sector CDOs,” Moody’s Investor Service, September 15, 2000 for a complete discussion of Moody’s methodology for computing the diversity score and application for rating CDOs.
20
and industry concentrations in the reference portfolio as well as default correlations across industries. Trading guidelines restricting the diversity score of a CDO’s portfolio to a minimum level are frequently used to prevent the portfolio from becoming overly concentrated in a single issuer or sector. A manager may not execute a trade that will result in a breach of the minimum diversity score restriction. In the same vein, individual issuer, sector, country and currency concentration limits are also frequently used to ensure that the collateral pool of assets remains diversified during the life of a CDO. Weighted Average Rating Factor (WARF) is a numerical metric used by Moody’s 8 to express the credit quality of a collateral pool of assets. It is derived by computing the weighted average of a numerical measure assigned to each rating category to reflect the expected defaults for that rating category. Exhibit 8 shows the values that Moody’s equates to each rating level. Averaging these values across the collateral pool, weighted by the par balance of the respective asset in the collateral pool results in the computation of the WARF statistic. The higher the WARF, the more likely the portfolio is to experience defaults. As with diversity score, trading guidelines restrict a collateral manager from adding or removing an asset from the portfolio if such an action would violate the maximum WARF test. exhibit 8
Moody’s Rating Factors
Aaa/AAA
Moody’s 1
Aa1/AA+
10
Aa2/AA
20
Aa3/AA-
40
A1/A+
70
A2/A
120
A3/A-
180
Baa1/BBB+
260
Baa2/BBB
360
Baa3/BBB-
610
Ba1/BB+
940
Ba2/BB
1,350
Ba3/BB-
1,780
B1/B+
2,220
B2/B
2,720
B3/B-
3,490
Source: Moody’s Investors Service
8
Moody’s interprets WARF as the expected 10-year cumulative default probability for a rating level. For example, for a B3 rating, 3,490 implies a cumulative default probability of 34.90% over 10 years.
Please see additional important disclosures at the end of this report.
21
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 1
Maximum Weighted Average Life (WAL) and Weighted Average Maturity (WAM) tests are intended to deter the collateral manager from taking duration bets inconsistent with the duration of the CDO liabilities. Other collateral quality tests include limitations on discount purchases – purchases of par assets with steeply discounted market value, payment-in-kind or deferred interest securities, and in the case of CLOs, the purchase of securities rated CCC/Caa. UNDERSTANDING THE BASICS OF CDO EQUITY RETURNS
The main motivation of a CDO equity investor is to achieve non-recourse term financing of the CDO’s underlying assets. The debt tranches used to fund the CDO assets are collateralized by the assets, and if the assets do not perform per expectations, the claims of the holders of the debt tranches are limited to the CDO assets. It is useful to compare this feature with the repo market, which is an alternative means of financing assets. In the repo market, the lender is not only collateralized by the pledge of the assets being financed but also has recourse to the borrower if the collateral is insufficient. With CDOs, the recourse is limited to the assets in the collateral pool of the CDO. Further, repo financing is relatively short term compared to the non-recourse financing achieved by equity tranche investors in cash CDOs. Effectively, equity investors seek to obtain a leveraged return as the positive difference between postdefault yield on the CDO’s assets and their cost of financing. Recall that the pricing of credit-risky assets compensates investors for expected credit losses and incorporates risk and liquidity premiums. The spread over the risk-free rate in an asset’s yield of a non-callable fixed-rate instrument is a measure of the reward for the credit risk and liquidity risk. As such, CDO equity investors are betting that the difference between the expected credit losses and experienced credit losses in the portfolio will be favorable and thus seek to capture risk and liquidity premiums in the assets in the collateral portfolio. In contrast to competing asset classes such as hedge funds, the cash flow return profile for CDO equity investors begins from deal inception. It is front-end loaded and not dependent upon the discretion of the collateral manager but on a predetermined set of rules. At deal inception, all structural protections and compliance mechanisms are in place. Therefore, for the first few years of the transaction, there may be residual cash flow available to equity tranche investors, which explains the front-loaded nature of CDO equity investments. For investors with exposure to first-loss, the likelihood of residual cash flows being available decreases over time as assets default. UNDERSTANDING THE BASICS OF CDO DEBT RETURNS
The risks to an investor with exposure to a portfolio of credit-risky assets may be thought of as having four dimensions – default probability, default severity or loss given default, default correlation and default timing. In a CDO, these risks are distributed to different tranches using the waterfall mechanism, with the debt tranches receiving the benefit of structural protection in the form of subordination, overcollateralization and coverage tests. Compared to an investor holding the assets in a collateral pool directly, an investor in a debt tranche of a CDO that holds the identical collateral pool of assets will
22
experience narrower loss default distributions because of the structural protections embedded in a CDO. The effects of diversification on the risks to investors are well known. Not only does a CDO offer the benefits of such diversification by requiring that the collateral pool of assets be diversified through the collateral quality tests but, in addition, the structural protections ensure that each additional default in the collateral pool have a small impact on the return of the debt tranche. The application of the CDO technology enables investors to gain exposure to asset classes they might not otherwise have access to. For example, investors mandated to limit investments to investment grade securities have very limited opportunities to invest in emerging market securities, few of which are rated investment grade. However, CDOs make it possible for such investors to get such exposure through investment grade tranches of CDOs with a collateral pool of emerging market securities. The same analogy is extendible to several other asset classes, such as leveraged and middle market loans. RATING OF CDOs
The market for the debt tranches of cash CDOs developed as a rated market and remains so even today. There have been several important methodological innovations in risk-neutral valuation approaches to structured credit, particularly the development of correlation models; 9 however, because of the many unique features of cash CDOs, their application to cash CDOs remains challenging. As such, ratings and their stability over time have remained important considerations for CDO investors. In this section, we briefly summarize the approaches used by the three major rating agencies – Moody’s, S&P and Fitch – for rating cash CDOs. While all three agencies base ratings on their own quantitative models, qualitative factors such as the experience of the collateral manager, the legal structure, and credit quality of the hedge counterparty, if any, also have a significant influence on the ultimate rating assigned to CDO tranches. Moody’s Moody’s ratings are based on an assessment of the probability that the collateral will generate sufficient cash flows to meet the obligations under each class of rated notes. Moody’s assessment is based largely on a statistical analysis of historical default rates on debt obligations with various ratings, the asset and interest rate coverage requirements for each class of notes and the diversification requirements the CDO is covenanted to satisfy. For a long time, Moody’s has used variations of the widely known binomial expansion technique (BET) for rating cash CDOs. Diversity score, a concept introduced by Moody’s, is at the heart of the BET approach. The BET approach maps the reference portfolio of assets into a hypothetical portfolio of N number of uncorrelated, homogeneous assets with identical default probabilities 10 and equal par values for calculating expected loss distributions, where N is the diversity score, discussed earlier. The formula for calculating the diversity score 9
See Chapter 63 for an analysis that contrasts rating agency approaches with risk-neutral valuation models. 10 The identical default probabilities are assumed to equal the weighted average default probability of assets in the original reference portfolio of assets.
Please see additional important disclosures at the end of this report.
23
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 1
incorporates obligor and industry concentrations in the reference portfolio as well as default correlations across industries. In the context of multi-sector CDOs, Moody’s recognizes that the loss severity of an asset varies by asset type, credit rating and position within the capital structure. For asset types that are prepayment sensitive, Moody’s takes into account maturity shortening or extension risk due to faster or slower prepayment speed. The BET model calculates expected losses stemming from defaults in the hypothetical portfolio, going from zero defaults to N defaults and assigning a probability to each default scenario. Moody’s then determines the level of subordination necessary to achieve the expected loss associated with the targeted rating for each of the rated tranches. A Moody’s rating represents an opinion on the expected loss of each tranche, expressed as the difference between the present values of the expected payments and the promised payments. We should note that the BET model is also used for rating cash CLOs. Moody’s introduced the Correlated Binomial modeling framework (CBM) in August 2004, which is now being used in modeling cash flow ABS CDOs. The CBM is very similar to the BET methodology in that the CBM also calculates expected losses from defaults in a representative portfolio. However, while the CBM incorporates the correlation among the representative assets as an explicit modeling parameter, BET assumes that the assets in the portfolio are independent. The single asset correlation parameter used in CBM (referred to as “Moody’s Asset Correlation”) is derived by using Moody’s Monte Carlo simulation-based CDOROM model. Standard and Poor’s (S&P) The S&P approach is different from the Moody’s approach in that the rating addresses the first dollar loss for a given rating category as opposed to mapping expected losses into a specific rating category. The S&P approach for rating cash CDOs is based on its proprietary model called the “CDO Evaluator.” The model takes into account the credit rating of each asset in the collateral pool, the number of assets in the pool, industry concentration and a default correlation between pairs of assets. The model uses Monte Carlo simulations to generate a probability distribution of defaults for each individual asset in the pool. The higher the targeted rating for a tranche, the higher the level of defaults the collateral pool of assets must withstand. The simulations take into account individual asset default probabilities based on rating and pair-wise correlations between the assets in the reference portfolio. The model draws a large number of multivariate normally distributed numbers, which are then compared with a default threshold based on the asset’s default probability and maturity to decide whether an asset defaults. Correlations in the Evaluator model are based on historically observed defaults, with asset correlation calibrated to default correlation. For cash CDOs, the S&P model explicitly incorporates intra-industry correlations, but treats inter-industry correlations as being zero. Last year, S&P announced that it will roll out a new version of the CDO Evaluator for the cash flow CDOs. The new cash flow model is likely to incorporate explicit assumptions for inter-industry correlations, as well. S&P classifies corporate industries at local, regional and global levels, and gives an additional credit for geographically diverse portfolios.
24
Fitch Fitch’s ratings are based on its Monte Carlo simulation platform, VECTOR, which was introduced in 2003. Like S&P, Fitch’s approach also addresses the probability of a first dollar loss. The model CDO portfolio default distributions are based on Monte Carlo simulations using individual asset default rates and asset correlations as the key inputs. The default rates are derived from a matrix that provides default rates by rating and maturity based on historical realized defaults. Pair-wise correlations are based on estimates of intra- and inter-industry correlations, as well as geographical correlations of equity returns. The underlying statistical model for estimating correlations is a factor model that expresses the return of the equity security of a firm as a function of statistically determined factors and company-specific idiosyncratic risk. Using this approach, average factor loadings and average idiosyncratic risk exposures are computed for each industry-geographic region grouping and used to derive asset correlations. ANALYTICAL CHALLENGES IN MODELING CASH CDOs
The many features that distinguish CDOs from competing asset classes – their adaptability to various asset classes, structural innovations and portfolio management strategies – render them challenging to model in analytically tractable and computationally convenient means. As a result, typical models in vogue for cash CDO analysis still rely on deterministic scenario analysis using annual constant default rates incorporating prepayment and recovery assumptions. Recently, there has been a gradual shift towards Monte Carlo simulation based analysis, which is a welcome improvement. The complex waterfall structures and the potential for diversion of cash flows because of structural protections and coverage triggers imply that analytical models have to take into account the default probabilities as well as the determinants of the different coverage ratios in a mutually consistent fashion. This, coupled with the optional redemption feature of most CDOs, calls for the modeling of interest rate risk in conjunction with the default risks of assets. Since the underlying collateral pools are managed, albeit according to predefined guidelines, the impact of future trading activities of collateral managers is difficult to model. Some analyses model the trading provisions assuming that the managers will trade at the extremes of all possible trading constraints, which has the effect of painting all managers with the same brush, constraining relative value judgment. Trading further complicates the already difficult problem of modeling and parameterization of correlation. Ultimately, it is important to keep in mind that analytical modeling needs to be done in conjunction with regular surveillance of the performance of both the underlying collateral and the CDO, in the context of the many structural mechanisms.
Please see additional important disclosures at the end of this report.
25
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 2
A Synthetic CDO Primer INTRODUCTION
Since their first appearance in 1997, synthetic Collateral Debt Obligations (CDOs) have gained enormous popularity and have been the most fertile area of growth and innovation in the structured credit markets, having contributed to as well as benefited from the explosive growth in the use of credit default swaps (CDS). From their initial application as a mechanism of risk transfer from bank balance sheets to manage regulatory capital requirements, they now encompass every facet of credit risk covering a wide range of assets from corporate bonds and loans, structured finance obligations and CDO tranches themselves. Given that the bulk of the synthetic CDO issuance takes place in private transactions and the amount of public data available is limited, we demonstrate the size of the synthetic CDO market by presenting two sets of data. Exhibit 1 shows the growth in the rated synthetic CDO notes over the period 1998-2005. Since the full capital structure of a synthetic CDO is typically not rated, this measurement of issuance represents only a fraction of the underlying portfolios and the embedded credit risk. Exhibit 2 is shows funded and unfunded deal volume as reported to Creditflux, a popular structured credit publication. These data represent an estimate of the credit risk that has been distributed using synthetic CDO technology to end investors, in both cash and synthetic form, and capture deals not included in the data on rated notes. These numbers compare favorably with the cash CDO issuance numbers discussed in Chapter 1. In fact, synthetic CDO issuance appears to have overtaken the cash CDO issuance. By either measure, the volume of synthetic CDO issuance and its recent rate of growth are impressive. Particularly noteworthy is that over half of the synthetic CDO market is denominated in non-USD terms, particularly in Euros. exhibit 1
Global Rated Synthetic CDO Issuance (USD Billions)
90 80
USD
EUR
Others
70 60 50 40 30 20 10
1998
Source: CreditFlux, S&P
1
26
1999
2000
2001
2002
2003
2004
1
For each year, end of the year exchange rates were used to convert non-USD issuance into USD.
2005
exhibit 2
Global Total Synthetic CDO Issuance (USD Billions)
$ Billion 350
Unfunded Funded
300 250 200 150 100 50 0 2002
2003
2004
2005
Source: CreditFlux
WHAT ARE SYNTHETIC CDOs?
Synthetic CDOs are the result of innovative combination of two technologies – the securitization techniques applied to transfer credit risk by cash CDOs and credit derivatives, which enable the isolation of credit risk from other components of risk. Effectively, synthetic CDOs are CDOs using CDS. Within that broad description, there is a wide variation in the underlying credit risky assets, whether or not they are managed, the extent of associated funded issuance and the motivation behind the risk transfer, each of which defines a particular type of synthetic CDO. To motivate a discussion of how synthetic CDOs work, we revisit the basic mechanics of a credit default swap (Exhibit 3). Recall that a CDS is akin to an insurance policy that protects the buyer of protection against the loss of principal in an underlying asset when a credit event occurs. The protection buyer pays a premium, typically on a quarterly basis to the protection seller until a credit event occurs or the contract matures, which ever is earlier. The underlying asset is defined by a reference obligation, which informs the scope of the protection. When a credit event occurs, depending upon the settlement mechanism specified in the CDS contract, the buyer of protection delivers a reference obligation to the seller and receives par in return (physical delivery) or receives the difference between the par amount of the reference obligation and its recovery from the seller (cash settle). Standard credit events include bankruptcy, failure to pay and restructuring of the debt. (See Chapter 1 of Credit Derivatives Insights – Single Name Instruments and Strategies, 2nd Edition, February 2006 for a complete description of CDS mechanics and the associated terminology).
Please see additional important disclosures at the end of this report.
27
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 2
exhibit 3
How a Typical CDS Works
Before a credit event occurs Premium Protection Seller
Protection Buyer After a credit event occurs (Physical settlement) Deliverable obligation
Protection Seller
Protection Buyer Par amount After a credit event occurs (Cash settlement)
Protection Seller
Par amount - recovery
Protection Buyer
Source: Morgan Stanley
Much like a cash CDO, a synthetic CDO typically involves a special purpose vehicle (SPV) that acquires exposure to a collateral pool of credit risky assets that is distributed to investors in a tranched form with different tranches having varying levels of credit risk and correspondingly varying levels of returns. Unlike cash CDOs in which the SPV purchases the assets from the market or a sponsoring financial institution, in synthetic CDOs the SPV acquires the risk exposure through credit derivatives, typically credit default swaps. The SPV sells protection on the collateral pool of assets to the sponsoring financial institution or other market participants and receives a premium for the risk being assumed. The credit risk so acquired is distributed to investors of different tranches who receive a portion of the premium depending upon the amount of credit risk assumed by each tranche. When a credit event occurs with respect to any asset in the collateral pool, the SPV pays the protection buyer an amount linked to the loss incurred on the asset. The loss is then passed on to investors in reverse order of seniority (i.e., the junior most tranche bears the first loss). Often, some of these tranches are funded tranches, analogous to credit linked notes with the only difference being that the risk and return are linked to a portfolio of assets as opposed to a single asset. The proceeds of the funded note issuance are invested in low-risk “eligible collateral”. 2 We elaborate further on funded and unfunded synthetic CDOs in a later section. Thus, in a synthetic CDO, investors act as the sellers of protection on a pool of underlying assets, the sponsoring financial institution or the market participants are the buyers of protection and the SPV is the intermediate vehicle that effectively distributes the cash flows involved. As such, synthetic CDOs are the derivative counterparts of cash CDOs.
2
Eligible collateral mainly consists of investments in cash or government bonds or guaranteed investment contracts (GICs) issued by highly rated insurance companies.
28
exhibit 4
How a Typical Synthetic CDO Works
Premium Credit Default Swaps with Sponsoring Financial Institution or Market
Premium
Super Senior
SPV (CDO Issuer) (Reference Pool of Assets) Protection
Protection Principal & Interest
Note Proceeds
Rated Notes Senior and Mezz Tranches Equity Tranche
Notes
Protection Buyer
Proceeds
Eligible Collateral
Excess Spread
Protection Seller
Source: Morgan Stanley
ATTACHMENT AND DETACHMENT POINTS
It is important to understand the terminology of attachment and detachment points to follow the mechanics of synthetic CDOs. An attachment point, expressed as a percentage or an absolute value, defines the amount of losses in the reference pool of assets that need to occur before a particular tranche starts to experience losses. A detachment point, also expressed as a percentage or an absolute value, defines an amount of losses in the reference pool of assets that need to occur for a complete loss of principal for that tranche. The size of each tranche (width) is the difference between the attachment and detachment points for that tranche and defines the maximum amount of losses that the tranche will experience. Exhibit 5 illustrates a possible capital structure of a hypothetical synthetic CDO with high yield unsecured corporate credit risk exposure. The numbers in parentheses are tranche widths. The equity tranche has an attachment point of 0% and detachment point of 10%. The first 10% of the credit losses from the underlying portfolio are absorbed by the equity tranche. Similarly, the super senior tranche has an attachment point of 30% and detachment point of 100%. All portfolio credit losses exceeding 30% are borne by the super senior tranche. If the portfolio experiences losses equal to 15%, the equity and the BB tranches would be wiped out and the BBB tranche would lose half of its notional.
Please see additional important disclosures at the end of this report.
29
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 2
exhibit 5
Capital Structure of a Hypothetical Synthetic CDO Losses from Reference Portfolio are borne by tranches in reverse seniority order
Super Senior Tranche (70%)
AAA Tranche (7.5%)
30% 22.5%
AA Tranche (3%) A Tranche (2.5%) BBB Tranche (4%) BB Tranche (3%) Equity Tranche (10%)
19.5% 17% 13% 10% 0%
Note: The ratings and the capital structure shown above are merely for illustration purposes and do not represent an actual transaction. Source: Morgan Stanley
WHY SYNTHETIC CDOs?
It is no coincidence that the surge in investor interest in synthetic CDOs has coincided with exploding trading volumes in credit default swaps and standardized index tranches. The application of credit derivatives technology has led investors with a wide range of interests to explore synthetic CDOs. In this section, we outline some of the key motivations for synthetic CDOs over their cash counterparts. Funding Gap and the Super Senior Tranches The larger the funding gap, 3 the higher is the potential return to the equity tranches for a given level of leverage. Expressed differently, higher funding gap would enable a transaction to be structured with lower leverage for a given level of target return for the equity tranches. The larger funding gap of synthetic CDOs is primarily due to the existence of the so-called unfunded super senior tranches, which receive a spread lower than the spread paid to the Aaa/AAA note tranches of cash CDOs. This lowers the average funding cost for the CDO (the weighted average spread paid to non-equity tranches). In essence, synthetic CDOs enable investors to efficiently capture risk premiums resulting in a ratings arbitrage. Increased Flexibility and Customization Relative to cash CDOs, structuring synthetic CDOs provide increased flexibility. Synthetic technology makes it possible to customize risk exposures, with respect to currency, cash flow, tenor and size of exposure. Since there is no need for the SPV to identify and acquire a specific asset, the maturity of the transaction is not constrained by the maturity of the pool of assets. With the increased liquidity of credit default swaps across different maturities, it is possible to customize synthetic CDOs to shorter 3
We define funding gap of a synthetic CDO as the difference between the weighted average spread/premiums received by the SPV and paid to the different debt tranches of the CDO.
30
maturities, an option typically unavailable with cash CDOs. It is also possible to structure the CDO liabilities such that the tranche coupon payment dates exactly match the premium receipt dates on the underlying CDS transactions. There is also significantly lower ramp-up period and therefore lower carry costs. The increased flexibility brings higher efficiencies and lower costs. Wide Range of Permissible Assets Synthetic CDOs enable managers to gain exposure to wide range of assets and strategies. For example, it is possible to structure CDOs to have long and short risk exposure whereas cash CDOs are typically long only. Efficiencies of Unfunded Liabilities Considering that there is not always a funded portion of a synthetic CDO, several sponsoring financial institutions find unfunded liabilities cheaper, easier and significantly faster to execute. A TAXONOMY OF SYNTHETIC CDOs
The innate flexibility of synthetic structures implies a wide variety of synthetic CDO types depending upon the motivation, funding, underlying risk exposure, collateral management strategies and liability structure. In this section, we illustrate some of the major categories of synthetic CDOs and discuss the unique issues involved with each of them. Balance Sheet Versus Arbitrage Obtaining economic and/or regulatory capital relief was the original motivation for the development of synthetic CDOs. Transactions so motivated are balance sheet synthetic CDOs in which a financial institution, typically a bank, uses a credit default swap to remove credit exposure to a portfolio of credit risky assets from its balance sheet while retaining its ownership of such assets. This enables the sponsoring financial institution to maintain lending relationships with the entities issuing the credit risky assets and at the same time achieve capital relief and efficient credit risk management. The portfolios underlying such transactions are typically static with little substitution privileges. In contrast, arbitrage synthetic CDOs, whose evolution followed that of the balance sheet genre, are designed to take advantage of the difference between the spread received from selling protection on individual reference assets/entities and the spread paid to investors to buy protection on a tranched basis. Unlike balance sheet CDOs, the credit risky assets are not on the sponsoring institution’s balance sheet. As with several cash CBOs, excess spread plays an important role in arbitrage synthetic CDOs, often used to offset losses or to hedge against future losses and in some cases returned to the investors of equity tranches. Single tranche CDOs are an increasingly popular type of arbitrage CDO. The prevalence of bespoke portfolios looking for effective risk management vehicles is probably the major motivation behind these structures that allow the investor to customize the credit risk exposure to assume or lay off by picking the reference pool as well as the attachment and detachment points.
Please see additional important disclosures at the end of this report.
31
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 2
Static Versus Managed The underlying portfolios in arbitrage synthetic CDOs can be either static or managed. In a static portfolio, the reference pool of entities is fixed at deal inception. In managed transactions, changes can be made to the reference portfolio by a designated collateral manager, within a broad framework of trading guidelines and restrictions. The trading guidelines and restrictions are designed to ensure that the collateral managers manage the risks and returns in the portfolio according a predetermined level of credit quality, diversification and risk management. There is a wide variation in the nature and scope of the trading guidelines. In some cases, the guidelines permit the manager to have long and short risk buckets allowing for the acquisition as well as hedging of credit risk subject to constraints. In dealing with the trading guidelines, investors and rating agencies have to contend with a significant dilemma. In general, trading out of a deteriorating credit may be thought of as a prudent risk management measure. However, trading out of a deteriorating credit also usually implies having to make termination payments and incur losses because of trading. Some transactions treat such losses as subordination erosion without replenishing the notional amount of the underlying portfolio. However, some transactions allow for the replenishment of notional traded out which introduces the potential for adverse selection that misaligns the interests of debt and equity tranches. Rating agencies impose maximum spread per reference credit limits to address the potential for such adverse selection, which is not unlike the limits placed in cash CDOs on the purchase of deeply discounted securities. An alternative approach to addressing this issue is through the excess spread mechanism. In many cases, the structure allows the excess spread to cushion the tranches from trading losses. Trading profits are used to supplement the excess spread. Trapping the excess spread until maturity or allow the leaking out of the excess spread only as long as the deal is performing satisfactorily are some other mechanisms used to address adverse selection issues. In many cases, trading ceases at some predefined point of the CDO life cycle or when trading losses reach a certain predefined level. Rating agencies review trading guidelines by focusing on isolating trading losses such that losses to investors in rated tranches come from credit events or risk management of credit risk in the underlying portfolio and not because of discretionary trading. Typically, the portfolios in the single tranche CDOs are static. Funded Versus Unfunded 4 Unfunded portfolio credit default swaps are the most basic of all synthetic CDOs (Exhibit 6). The protection buyer enters into a CDS on a specific portfolio of reference entities with the SPV, which in turn obtains protection by entering into CDS on a tranched basis with the investors, the ultimate sellers of protection. The CDS with investor(s) in each tranche defines the attachment and detachment points and the credit events. As sellers of protection, each tranche receives a fixed spread applied to the tranche size. As losses occur, the sellers of protection make loss payments to the SPV in the reverse order of seniority, passed on to the protection buyer.
4
The diagrammatic representations used in this section have been adapted from “Moody’s Approach to Rating Synthetic CDOs,” Moody’s Investor Service, July 29, 2003.
32
The structure of the funded synthetic CDO is more complicated (Exhibit 7). In funded tranches, the CDO investor pays the notional amount of the tranche at deal inception and losses due to any credit events result in principal writedown. Throughout the life of the transaction, the investor receives LIBOR/EURIBOR plus a spread, which reflects the riskiness of the tranche. The amounts paid by tranche investors (the proceeds of issuance) are invested in eligible collateral; typically, either GICs issued by highly rated insurance companies or highly rated short-term securities. The maturity dates of the GICs are set to match the maturity dates of the funded notes. Premium income from the CDS written by the SPV and interest income received by investing the proceeds in eligible collateral form the source of income to the SPV used to pay interest to the funded note holders and premiums to the unfunded tranches. CDS
exhibit 6
Structure of a Typical Unfunded Synthetic CDO
CDS
Credit Protection Sponsor or the Market
Premiums CDO Investors
SPV (CDO Issuer)
Premiums Ultimate Protection Buyers
Credit Protection Ultimate Protection Seller s
Reference Portfolio of Assets
Source: Morgan Stanley
exhibit 7
Eligible Collateral (GICs etc)
Structure of a Typical Funded Synthetic CDO
CDS
Note Proceeds
Coupon and Contingent Prinicipal
Credit Protection
Sponsor or the Market
SPV (CDO Issuer)
Coupon and Contingent Prinicipal
Note Proceeds
Funded Tranche Investors
Premiums Premiums
Credit Protection
Ultimate Protection Buyers
Reference Portfolio of Assets
Super Senior Tranche
Ultimate Protection Sellers
Source: Morgan Stanley
In practice, most funded synthetic CDOs are actually partially funded, in that only parts of the CDO capital structure is represented by funded tranches. Super senior and equity tranches are typically unfunded.
Please see additional important disclosures at the end of this report.
33
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 2
RATING AGENCY APPROACHES TO SYNTHETIC CDOs
While not all synthetic CDO tranches are necessarily rated, 5 ratings are important benchmarks for assessing investments in CDO tranches. In this section, we provide a broad overview of the rating methodologies used by the three major rating agencies. In general, the three approaches involve attributing default probabilities to each reference entity in the reference pool of assets and assumptions regarding the correlation of default probabilities and recovery rates. Incorporating correlation formally into rating agency models is a relatively recent phenomenon. The three major rating agencies have developed their own distinct methodologies for rating synthetic CDOs.6 For each agency, the basic framework for cash and synthetic CDOs is similar. However, certain stress factors and haircuts are applied in analyzing synthetic CDOs to take several unique features 7 of credit default swaps into consideration. It is important to point out that analytical models are only a part of the rating agencies’ analyses and each agency relies on other qualitative considerations such as deal documentation, legal assessment and evaluation of all parties involved in a transaction such as the collateral manager, hedge counterparty, servicer to arrive at a specific rating for each tranche. Moody’s A Moody’s rating represents an opinion on the expected loss of each tranche, expressed as the difference between the present values of the expected payments and the promised payments. For a long time, Moody’s used variations of the widely known binomial expansion technique (BET), first developed in the context of cash CDOs, for rating synthetic CDOs. In 2004, Moody’s introduced its new Monte Carlo simulation based models called CDOROM™ for rating synthetic CDOs marking a significant departure from its previous approach. The BET approach maps the reference portfolio of assets into a hypothetical portfolio of N number of uncorrelated, homogeneous assets with identical default probabilities 8 and equal par values for calculating expected loss distributions. The number N in the hypothetical portfolio is the well known “diversity score”, which is a measure of portfolio diversification – the higher the diversity score, the more diversified the reference portfolio. The formula for calculating the diversity score incorporates obligor and industry concentrations in the reference portfolio as well as default correlations across industries. The model calculates expected losses stemming from defaults in the hypothetical portfolio, going from zero defaults to N defaults and assigning a probability to each default scenario. Calculating probability-weighted losses for each CDO tranche results in the expected loss for that tranche, which is then mapped into a specific rating.
5
Funded tranches of synthetic CDOs are generally rated. The “super senior” and the equity tranches are typically not rated. 6 For complete discussion see “Moody’s Approach to Rating Synthetic CDOs”, Moody’s Investors Service, July 29, 2003; “Criteria for Rating Synthetic CDO Transactions”, Standard and Poor’s, September 2003 and “Global Rating Criteria for Collateralized Debt Obligations”, Fitch Ratings, September 13, 2004. 7 Some of these unique features include the counterparty credit risk, soft credit event, and settlement mechanics. 8 The identical default probabilities are assumed to equal the weighted average default probability of assets in the original reference portfolio of assets.
34
Over time, Moody’s has refined the BET approach by the use of the multiple binomial variation for rating synthetic CDOs which further divides the reference pool of assets into sub-pools and models the default behavior of each sub-pool with its own binomial analysis. Moody’s latest model, CDOROM™, is a Monte Carlo simulation based model that models the default behavior of the assets in the reference portfolio and explicitly incorporates both intra and inter industry correlations. The model also simulates correlated recoveries to take into account systematic variation in recoveries. Moody’s addresses the soft credit events by applying a stress factor to the default probabilities used to model transactions. A stress factor of 12.5% is applied for transactions using the 1999 ISDA definitions and include restructuring credit event without any supplements. A stress factor of 5% is used for transactions that use the 2003 ISDA definitions or if all the supplements are applied to the 1999 definitions. Recovery rate haircuts are applied to take into consideration the cheapest-to-delivery option embedded in the settlement of credit default swap upon a credit event. Moody’s applies a haircut of 5% and 10%, respectively, for investment grade and below investment grade assets, if there is no restructuring credit event, or restructuring credit event with restructuring maturity limitation, or modified restructuring with restructuring maturity limitation. The haircuts applied increase to 10% and 15%, when transactions include restructuring credit event without restructuring maturity limitation or modified restructuring maturity limitation. Standard and Poor’s (S&P) The S&P approach is based on the CDO Evaluator. Introduced in 2001, it was the first rating agency model to apply Monte Carlo simulation methods for rating CDOs. The S&P approach is different from the Moody’s approach in that the rating addresses the first dollar loss for a given rating category as opposed to mapping expected losses into a specific rating category. The simulations take into account individual asset default probabilities based on its rating and pair-wise correlations between the assets in the reference portfolio. The model draws a large number of multivariate normally distributed numbers, which are then compared with a default threshold based on the asset’s default probability and maturity to decide whether an asset defaults. The model produces a correlation adjusted probability distribution of potential, aggregate default rates for the collateral pool of assets. As such, it may be seen as an estimate of the distribution of aggregate defaults and losses at different probability levels. S&P made some meaningful changes in its CDO Evaluator model in December 2005 to bring its modeling assumptions in line with the new historical transition and default data – S&P CDO Evaluator v3 and v3.1 include these changes. In the new version, default probabilities for corporates are generally lower for investment grade collateral and higher for high yield collateral. The term structure of corporate default probabilities is also significantly different. For higher ratings categories within investment grade, the default probability curves are much flatter. On the other hand, for most below investment grade categories, the default curve is steeper in the new version. Correlations in the Evaluator model are based on historically observed defaults, with asset correlation calibrated to default correlation. S&P classifies corporate industries at local, regional and global levels and gives an additional credit for geographically
Please see additional important disclosures at the end of this report.
35
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 2
diverse portfolios. The new version of the CDO Evaluator explicitly incorporates both intra- and inter- industry correlations. In the past versions, inter-industry correlations were treated as zero. To account for the unique features of CDS, S&P applies haircuts to the base-case recovery assumptions used in the CDO Evaluator. A hair cut of 5% is applied to the base-case recovery assumption to address the cheapest-to-deliver option. Likewise, a hair cut of 2.5% each is applied to account for the specified currency settlement feature, if consent required loans are deliverable and when the period between the credit event notification date and the valuation date is deemed too short (S&P prefers transactions to have the longest possible valuation period, with 45 days as the minimum). A haircut of 10% is applied when a transaction allows for the so-called old restructuring as a credit event (1999 ISDA definitions without any supplements). A 21% haircut is applied in transactions where the deliverable obligation is denominated in a currency other than the floating payment currency. All the haircuts discussed are mutually exclusive. Fitch Fitch’s ratings are based on its Monte Carlo simulation platform, VECTOR, which was introduced in 2003. Like S&P, the Fitch approach also addresses the probability of a first dollar loss. The model CDO portfolio default distributions on the basis of Monte Carlo simulations using individual asset default rates and asset correlations as the key inputs. The default rates are derived from a matrix that provides default rates by rating and maturity based on historical realized defaults. Pair-wise correlations are based on estimates of intra and inter industry as well as geographical correlations of equity returns. The underlying statistical model for estimating correlations is a factor model that expresses the return of the equity security of a firm as a function of statistically determined factors and company specific idiosyncratic risk. Using this approach, average factor loadings and average idiosyncratic risk exposures are computed for each industry-geographic region grouping and used to derive asset correlations. ANALYTICAL CHALLENGES IN MODELING SYNTHETIC CDOs
While synthetic CDOs are significantly simpler than their cash counterparts in terms of valuation because of a significantly simpler waterfall, cash flow diversion rules and optional redemption potential, they still pose significant analytical challenges. While this is true both under a risk neutral valuation framework as well as objective or historical probabilities, the challenges are best understood juxtaposed against the former. In this section, we explore some of these challenges. The application of Gaussian copula models for analyzing correlated defaults in standardized credit index tranches has become an industry benchmark in much the same manner as the Black-Scholes model has been for equity option pricing (the technical details are explored in greater detail in Chapter 5). In fact, standardized credit index tranches can be thought of as specialized cases of synthetic CDOs. The differences stem from mainly two sources, which we explore in this section. First, the underlying portfolios can be managed or static pools with the index and second, there is an additional layer of counterparty credit risk with synthetic CDOs.
36
Portfolios that allow trading pose several analytical challenges for valuation and risk management of synthetic CDOs. With static portfolios, knowledge of the credit quality of reference entities in the underlying portfolios, their correlation, the CDO capital structure and the cash flow waterfall structure of the transaction are sufficient to make a reasonable determination of the credit risk and returns for a given tranche. With managed portfolios, since the underlying reference entities might be changing, the problem is more complicated. How does an analyst determine the credit quality of reference portfolios or calibrate the appropriate correlation levels when the constituents of the portfolios are unknown? One approach is to assume that the managers will always trade at the limit of the trading constraints. For example, if there is a maximum spread constraint, assume that the manager will always trade at that spread level. While some might consider this a prudent approach, it is not necessarily adequately reflective of the risks and returns an investor in a given tranche faces and has the effect of treating all managers with the same broad brush stroke. In a similar vein, correlations can be implied from observed prices of standardized index tranches. Estimating the same for non-standard synthetic CDO tranches with managed underlying portfolios imposes an additional layer of constraints. There are two layers of counterparty credit risk with synthetic CDOs. The SPV enters into CDS with the sponsoring financial institution or the market in general and purchases credit protection from investors in funded and unfunded tranches. This introduces credit risk in transactions outside of the credit risk contained within the reference portfolio assets to the CDS counterparties. With funded tranches, there is also credit risk associated with the eligible collateral investments. NEW DEVELOPMENTS IN SYNTHETIC CDOs
The original synthetic CDO market was a largely a bespoke one that certainly needed boosts in liquidity and transparency to make it more viable and mainstream. This benchmark liquidity was not only important for visibility purposes, but it also has helped the bespoke market in important ways. The resulting flood of structural innovation today is both interesting and rather complicated. The key question is how useful (or not) the innovation really is, and the answers largely depend on who (in the investment community) is asking. In many ways, the deluge of innovation reflects the diversity of credit investors in the market and their demand for sophistication, which we would argue is the important thematic shift in the market. In Exhibit 8, we summarize some of the key innovative features that have either become well established recently, or have been just introduced and show (in our view) some promise. We would argue that in many cases, new features are actually a reaction to some type of risk that was exposed previously, such as excessive idiosyncratic risk during the last credit cycle or associated low recoveries. But others are strictly innovative, including things like interest rate hybrids and principal protection techniques borrowed from other markets.
Please see additional important disclosures at the end of this report.
37
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 2
exhibit 8
Key Innovations, or Those That Show Promise
Key Innovation Hybrid CDO Squared (w/ABS)
Our Thoughts Reaction to the over-levered CDO-squareds of several years ago. The large ABS portion serves to delever the structure
Pure CDO-Squared w/Thicker Tranches
Similar idea to yesterday’s CDO-squareds, but thicker tranches serve to delever the structure
Large Collateral Pools w/Less Overlap Investors have recognized the importance of overlap in CDOsquareds, particularly for subordinate tranches in the outer CDO. Lower levels of overlap are now common Fixed Recoveries
A reaction to the low recoveries from fallen angel defaults in 2001-2002. Fixed recoveries reduce uncertainty, particularly in mezzanine tranches, but they are a double-edged sword as IG recoveries can be high as well
Dynamic Principal Protection (CPPI)
Application of an age-old technique to credit for investors seeking principal protection. Long/short tranche strategies or hedge fund investments seems to be the natural application
Cross-Subordination
An interesting concept that allows subordination within inner tranches of CDO-squareds to be shared
Interest Rate Hybrids
A natural progression as funded credit investors demand differing exposure to interest rate risk, beyond strict fixed or floating rate instruments
Adjustable Subordination (or Adjustable Coupon)
A viable technique for meeting investor demand for managed structures while allowing deal arrangers to hedge risks
Source: Morgan Stanley
CDO SQUARED
The so-called CDO squared structures are the arena where many of the recent innovation have come to light. A CDO squared transaction is a CDO backed mainly and some times only by other CDO tranches. In effect, they are really CDOs of CDOs (Exhibit 9). They have the benefit that they are insulated from a first loss exposure in the underlying portfolios and thus removed from idiosyncratic risk phenomenon. On the other hand, they have the potential for concentration risk across different portfolios that are hard to measure, control or hedge. CDO SQUAREDS – A BRIEF HISTORY LESSON
Synthetic CDO-squared transactions have become popular in the market for a variety of reasons, and can play some interesting strategic roles in credit portfolios (for some of our earlier thoughts, see Chapters 11 and 63). From the perspective of innovation in the marketplace, they deserve special attention because of their relative size and the story behind them. Early CDO-squared transactions were quite different from today’s state-of-the art technology, demonstrating the learning process that market participants went through. The first deals were issued in 1999 and were generally managed cash flow structures comprised of a large number (80 or so) of already managed junior and senior mezzanine tranches of high yield CBOs, and to some degree CLOs. In retrospect, they were highly levered transactions with a fair amount of concentrated (overlapping) single-name exposure. Most performed poorly when the credit cycle turned and taught investors some important lessons.
38
exhibit 9
Anatomy of a CDO Squared
CDO-Squared Structure
Senior Notes
Collateral Corporate Credit and/or Asset Backed Securities Corporate Credit and/or Asset Backed Securities
Mezzanine Notes
Equity
Secondary CDO Tranches
Corporate Credit and/or Asset Backed Securities Corporate Credit and/or Asset Backed Securities
Source: Morgan Stanley
Today’s CDO-squared structures are different along several dimensions. First, they are generally static synthetic structures. Second, the inner CDOs generally do not comprise 100% of the portfolio; there is often a significant portion of ‘funding’ in the structure, usually through the form of AAA-rated ABS collateral, which serves to delever the structure. As a result, ‘first loss’ tranches in the outer CDO generally can carry investment grade ratings. Third, the inner CDOs tend to be senior mezzanine or junior senior tranches of investment grade portfolios. Finally, the transactions are relatively transparent, meaning that investors have complete information on credit exposure and credit overlap (which may not have been the case in the past). It should be noted that CDO-squared transactions where the inner CDOs make up 100% of the collateral do still exist, but typically they include relatively thick tranches (e.g., 10% for investment grade portfolios). Also, in the cash CDO markets, pure CDO squared transactions are reemerging in new forms including those comprised of leveraged loan CLO tranches. WHAT’S THE MOTIVATION? SYSTEMIC VS. IDIOSYNCRATIC RISK
Most investors who are attracted to CDO-squared structures are not looking to take on idiosyncratic risk, as there are better ways of doing that. Yet early transactions had plenty of idiosyncratic risk because inner CDO attachment points were low (on high yield portfolios) and overlap was relatively high. Today’s structures reflect a flight from idiosyncratic exposure, but leverage is still an important part of the trade. While even the first-loss tranches of CDO-squareds are far from default, they are levered plays on systemic risk. Investors are effectively selling credit convexity, much like those who sell deep out-of-the-money options. SURVIVAL OF THE FITTEST
Financial engineering in today’s credit markets reflects Darwinian theories, in our view. The market will ultimately be able to distinguish good ideas from bad, and only the former will survive, but it will take some work to get there. We are strong supporters of continuing structural evolution, especially when it results from previous experience, but we advise investors to maintain a discerning approach.
Please see additional important disclosures at the end of this report.
39
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 3
A Tranched Indices Primer While the market for outstanding synthetic CDOs is vast and quite diverse with respect to structural complexity, it remains fragmented, as we discussed in the introduction to this book. The emergence of standardized, tradable, and relatively diverse default swap indices in 2003 provided the opportunity for an on-the-run market for tranches to develop shortly thereafter. Today, the tranched index instruments are by far the largest liquidity point in the structured credit market, serve as an entry point for many investors and have been responsible for making the business mainstream. Within the context of today’s synthetic CDOs, the tranched indices are relatively simple instruments, and are effectively standardized static synthetic CDOs. Yet, they deserve special attention given the size of the market and the importance of understanding their subtleties. In this chapter, we discuss the basic mechanics of the standardized indices of CDS and then go on to elaborate on details of tranched risk on these indices. We believe it is important to understand the nuances of these products before getting into an in-depth discussion of correlation and greeks of tranched structures, which we cover in Chapters 5 and 6. Our discussion in this chapter applies largely to corporate credit indices and their tranches. Some of the details for emerging markets indices may be significantly different.
40
Credit Default Swap Indices Credit default swap indices are simply portfolios of single name default swaps, serving both as trading vehicles and as barometers of the market activity. While intuitively very simple, the indices are responsible for increased liquidity and popularity of tranching of credit risk. By buying protection on an index, an investor is protected against defaults in the underlying portfolio. In return, the buyer makes quarterly premium payments to the protection seller. If there is a default, the protection seller pays par in exchange for the reference obligation to the protection buyer. Exhibit 1 shows the cashflows in an index: exhibit 1
DJ CDX Investment Grade Index
CDS CDS CDS CDS CDS CDS CDS CDS CDS CDS CDS CDS CDS CDS CDS
Cash Flows Prior to Maturity/Credit Event Quarterly Premium
Protection Buyer
CDS CDS CDS CDS CDS CDS CDS CDS CDS CDS
Protection Seller
CDS CDS CDS CDS CDS CDS CDS CDS CDS CDS
Protection against credit event
125 equally weighted names Cash Flows in Case of a Credit Event
CDS CDS CDS CDS CDS CDS CDS CDS CDS CDS
Delivery of Senior Unsecured Obligation
Protection Buyer
CDS CDS CDS CDS CDS
Protection Seller Par
CDS CDS CDS CDS CDS CDS CDS CDS CDS CDS CDS CDS CDS CDS CDS CDS CDS CDS CDS CDS
Source: Morgan Stanley
Several investment grade and high yield indices trade in the US, covering multiple maturities, sub-sectors, credit quality, etc. In addition, there are numerous similar indices for Europe, Pan Asia, and Emerging Markets. Pricing levels, descriptions, and calculators for these indices are available on the MSCD
screen of Bloomberg. IMPORTANT CHARACTERISTICS OF BENCHMARK INDICES
Static Underlying Portfolio Once an index composition is fixed, no names can be added or deleted from the portfolio. It is also noteworthy that all names are typically equally weighted, as opposed to market weighted, which is common for benchmark bond indices. Rolling Over of Indices As time passes the maturity term of indices decreases, making them significantly shorter than the benchmark terms, so new indices are introduced periodically and the latest series of the index represents the current on-the-run index. Markets have continued to trade previous series of indices albeit with somewhat less liquidity.
Please see additional important disclosures at the end of this report.
41
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 3
Standardized Payment and Maturity Dates Just like the single name default swaps, the cash flow dates of indices are also standardized – the 20th of March, June, September, and December of every year. Market participants have also standardized maturity dates to the four standard payment dates of the maturity year. Deal Spread The indices have a predetermined “Deal Spread”, which is paid on a quarterly basis. Consequently, if the index is currently trading away from the deal spread, an upfront payment is required to reflect the difference between the current market spread level and the deal spread. Conceptually, it is equal to the present value of the difference between the two, adjusted for default probabilities. It is also important to note that all the underlying single name contracts also have the same deal spread as the index. Just as a portfolio of bonds with different coupons has better convexity than a corresponding portfolio with the same coupon for each of the bonds (assuming both portfolios have the same average coupon and maturity), the convexity characteristics of the index are somewhat different from that of an equalweighted portfolio of the underlying single name default swaps. Payment of Accrued Premiums If an investor enters an index transaction in between the payment dates, the protection seller would make a payment of accrued premium to the protection buyer, to reflect the fact that the protection buyer would pay premium for the full quarter on the next payment date but the protection is in effect only for part of the quarter. Restructuring Definitions The market standards regarding restructuring definitions for indices and underlying credit default swaps are not always the same. For example, while most of the underlying single names for DJ CDX NA IG index trade with a Modified Restructuring (Mod-R) definition, the index itself trades on a No-R basis. European indices, however, trade with the same restructuring definition as the underlying, Modified Modified Restructuring (Mod-Mod R). For further details on restructuring definitions, refer to Credit Derivatives Insights – Single Name Instruments and Strategies, 2nd Edition, February 2006. DETERMINING THE UPFRONT PAYMENT
As we mentioned earlier, if the index is trading away from the deal spread, an upfront payment is required to reflect the difference between the current market spread level and the deal spread. Theoretically, the present value of the two premium streams should match when we take default probabilities and timing of cashflows into consideration. The first step for calculating the upfront payment is to estimate default probabilities from the credit curve (for more details, refer to Credit Derivative Insights - Single Name Instruments & Strategies, 2nd Edition, February 2006). Using these probabilities, we calculate the present value of the current spread, by multiplying the spread with the probability of survival at the time of payment and then discounting back using risk-free zero rates. Now, this present value should equal the present value of upfront and running premiums (the Deal Spread), based on the same default probabilities. So if the
42
deal spread is higher than the current par spread, the protection seller makes a payment to the protection buyer. A convenient way to do this conversion is to use the CDSW function on Bloomberg. We simply put in the “Deal Spread” and value the contract using the current par spread. The “Market Value” represents the equivalent upfront payment. In addition to the upfront calculation, we can use this function to calculate mark-to-market, DV01 and cashflows. The DV01 is especially helpful in delta hedging of portfolio credit exposures using indices. IMPACT OF DEFAULTS ON INDEX CASHFLOWS
When an underlying single name defaults, it is separated from the index and settled separately. For example, for DJ CDX NA IG index, which has 125 names, if one of the underlying names defaults, the remaining index would have 124 names and the same deal spread. The 1/125th of the notional would be separated and the protection seller would pay par to the protection buyer in exchange of a deliverable obligation. After a default, the premium payments for the index would be (124/125)*Deal Spread, irrespective of which of the 125 names defaults (this methodology applies to CDX series 3, and may not hold for previous indices). This is due to the same deal spread for all underlying names in the index portfolio, as we mentioned earlier. It is important to note that an equal-weighted portfolio of underlying names could now have a different spread, given that the each of the underlying names has its own spread level and depending on which of the 125 names defaults, the average spread for the remaining 124 names could be different from 124/125 of the original spread. exhibit 2
Impact of Default on Index
CDS CDS CDS CDS CDS
CDS CDS CDS CDS
CDS CDS CDS CDS CDS
CDS CDS CDS CDS CDS
CDS CDS CDS CDS CDS
CDS CDS CDS CDS CDS
CDS CDS CDS CDS CDS
CDS CDS CDS CDS CDS
CDS CDS CDS CDS CDS
CDS CDS CDS CDS CDS
CDS CDS CDS CDS CDS
CDS CDS CDS CDS CDS
CDS CDS CDS CDS CDS
CDS CDS CDS CDS CDS
125 equally weighted names CDS CDS CDS CDS CDS
CDS
Defaulted name settled separately
124 equally weighted names CDS CDS CDS CDS CDS
CDS CDS CDS CDS CDS
CDS CDS CDS CDS CDS
CDS CDS CDS CDS CDS
CDS CDS CDS CDS CDS
CDS CDS CDS CDS CDS
CDS CDS CDS CDS CDS
CDS CDS CDS CDS CDS
CDS CDS CDS CDS CDS
CDS CDS CDS CDS CDS
CDS CDS CDS CDS CDS
CDS CDS CDS CDS CDS
CDS CDS CDS CDS CDS
Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
43
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 3
Tranches of Standard Default Swap Indices Co-existence of synthetic CDOs and liquid benchmark indices resulted in the logical next step of applying the tranching techniques to the indices, which lead to standardized synthetic CDOs. In this section we will discuss the basic structure of these instruments, followed by a more detailed analysis of their characteristics. THE BASIC MECHANICS
Borrowing the CDO technology, we can tranche the credit risk of an index into a number of slices, with different levels of subordination. The junior most tranches covers initial defaults, and once losses exceed the notional of the tranche they are passed on to the next senior tranche in the capital structure. For example, the most liquid standardized IG DJ CDX NA index tranches are 0-3%, 37%, 7-10%, 10-15%, and 15-30% (see Exhibit 3). The first tranche, also referred to as the “equity” tranche, takes the first 3% of the losses on the portfolio. When the portfolio has accumulated large enough losses to exceed 3% of notional, the next tranche, 3-7%, will incur losses from any potential further defaults, and so on. The standardized tranches for the high yield index, DJ CDX NA HY, are 0-10%, 10-15%, 15-25%, 25-35%, and 35-100%. How are the standardized tranches determined? The attachment and detachment points of standardized tranches are partly driven by the synthetic CDO market and expected ratings of the tranches. It is important to note above that the IG equity tranche takes 3% of overall portfolio losses, not defaults. Assuming a 40% recovery, 5% of the portfolio would have to default in order to wipe out the entire equity tranche.
Copyright 2006 Morgan Stanley
44
IMPORTANT CHARACTERISTICS OF BENCHMARK TRANCHES
Attachment and Detachment Points The attachment point determines the subordination of a tranche. For example, a 3-7% tranche’s attachment point of 3% implies that the tranche will incur losses only after the first 3% of the notional has been lost due to defaults over the term of the index. The detachment point determines the point beyond which the tranche has lost its complete notional. In other words, the 3-7% tranche is completely wiped out if portfolio losses exceed 7% of the index. The difference between the attachment and detachment points is referred to as “thickness” of tranche. exhibit 3
Tranched DJ CDX NA IG
IG DJ CDX NA Index CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
Tranched DJ CDX
Super Senior 30-100%
125 equally w td nam es CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
CDS
Junior Super Senior 15-30% Senior 10-15% Senior Mezz 7-10% Junior Mezz 3--7% Equity or First Loss 0-3%
Source: Morgan Stanley
Market Quoting Convention Tranches that are expected to incur significant default losses, i.e. the equity tranche and some other junior tranches, are typically quoted with a sizeable upfront payment and some running premium, while more senior tranches are typically quoted on a running premium basis. For investment grade (DJ CDX NA IG), the equity tranche is currently priced on an upfront plus 500 bps running, while all other tranches are priced on a running premium basis. However, in case of high yield (DJ CDX NA HY), the first two tranches are currently priced on an upfront basis with 0 bps running, while the more senior tranches are priced only on a running premium basis. Typically, when a tranche is expected to face material losses, an upfront payment helps reduce the volatility of returns and sensitivity of the tranche to spread changes. The reason being that in case of high defaults the running premium would drop significantly along with the outstanding notional, implying lower expected premiums. Conversely the protection seller collects higher premiums in a low default case. For more details on this topic refer to Chapter 6.
Please see additional important disclosures at the end of this report.
45
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 3
Tranche “Delta” The delta of a tranche reflects its sensitivity to changes in spreads of the underlying index portfolio. It is typically measured as a ratio of PV01 of the tranche to PV01 of the index. (PV01 is the change in the mark-to-market for a 1 basis point move in each of the underlying credits in the portfolio.) Typically, dealer quotes assume that the counterparty will also enter into a delta-neutral notional amount (i.e., delta x tranche notional) of the index simultaneously with a tranche transaction. Broadly speaking, junior tranches have higher deltas than senior tranches, if both are quoted on a running premium basis. However, the presence of upfront payments for junior tranches (equity tranche for investment grade DJ CDX and the first two tranches for high yield DJ CDX) lowers tranche delta. We will discuss the intuition and further details on this topic in Chapter 6. Rolling over of Indices As time passes the maturity term of tranches and the underlying index decreases, making them significantly shorter than the benchmark terms. So, tranches on the new on-the-run indices are introduced, as the indices roll over. Standardized Maturity Dates Market participants have standardized maturity dates for tranches to the maturity dates of the underlying indices. Standardized Payment Dates Just like the underlying indices, the cash flow dates of index tranches are also standardized – the 20th of March, June, September and December of every year. Payment of Accrued Premiums If an investor enters an index tranche transaction in between the payment dates, the protection seller would make a payment of accrued premium to the protection buyer, to reflect the fact that the protection buyer would pay premium for the full quarter on the next payment date but the protection is in effect only for part of the quarter. Restructuring Definitions The restructuring definition market standards for index tranches matches the underlying index convention (No-R for US indices).
46
Hybrid Settlement in Case of Default When one of the names in the index defaults, the loss is transferred to the relevant tranche (say, the equity tranche assuming that this is the first default in the portfolio). Determining the realized recovery on the defaulted name is very important, as it not only impacts the cashflows of affected tranches but also determines the erosion of subordination for other tranches. To determine the recovery rates more accurately, the market has adopted a “hybrid settlement” for index tranches. The goal of hybrid settlement is to balance the protection seller's desire for a high valuation with the protection buyer's interest in keeping recoveries low. Hybrid settlement has two stages: partial physical settlement followed by a cash settlement valuation. The first phase consists of obligations being delivered up to the amount of the defaulted name’s delta in each tranche. The second phase involves an auction process, which determines the amount by which each tranche is written down. 1
1
See Chapter 6 of Credit Derivatives Insights – Single Name Instruments and Strategies, 2nd Edition, February 2006, for more details on the ISDA protocol for settling index tranches.
Please see additional important disclosures at the end of this report.
47
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 4
A First-to-Default Basket Primer While the lion’s share of attention and liquidity in the tranched credit markets is focused on cash CDOs, synthetic CDOs and the tranched indices (see Chapters 1-3), first-todefault baskets, or more generally, Nth-to-default baskets are the simplest tranched credit instruments. First-to-default baskets are optically very similar to a credit default swap, but the uncertainty associated with whether a credit in the basket will trigger a credit event makes their risk and return profile intuitively more similar to CDO tranches. There are important differences, as well, which we address in this chapter. OVERVIEW OF BASKET DEFAULT SWAP MECHANICS
An Nth-to-default basket is a product with which the investor gains either long or short exposure to a relatively small basket of credits. The typical size of baskets ranges from 4 to 10 credits. The investor is either selling or buying protection on the Nth credit to default in the basket, as with a plain vanilla credit default swap, except that the reference credit is a basket instead of a single name. Five-name baskets are most common, and protection is typically bought or sold on either First-to-Default (FTD) or 2nd-through-5th default basis. Suppose an investor sells FTD protection on a five-name basket of reference credits. The investor will receive a periodic payment (the “premium”) in exchange for taking on the credit risk of the basket. If no credit events occur during the term of the basket default swap, the swap expires. If a credit event occurs during the term of the basket, the swap is terminated and the “buyer” of protection delivers the reference credit that experienced the credit event to the “seller” of protection in exchange for a par payment (equal to the notional amount of the swap). This process is analogous to a single name credit default swap. The settlement in case of default can be done either through a cash settlement or physically, with a security that qualifies for delivery under the terms of the underlying default swap on the credit. A credit event is defined as it is in the single name market, with investors having the option of choosing the restructuring definition. When a credit event occurs, the seller of protection in the basket is effectively losing par minus the recovery value of the defaulted asset. For detailed definitions of credit events and deliverable obligations, refer to our publication, Credit Derivatives Insights – Single Name Instruments and Strategies, 2nd Edition, February 2006. The following exhibit shows an FTD swap with an underlying basket of five names:
48
exhibit 1
Name 1
First-to-Default Swap Name 2
Name 3
Name 4
Name 5
2ndthrough5th Default
First-toDefault
Source: Morgan Stanley
SELLING FIRST-TO-DEFAULT PROTECTION: LOOKS LIKE A DEFAULT SWAP, FEELS LIKE CDO EQUITY
Although basket default swaps resemble single-name credit default swaps optically, their risk/return and valuation profiles are much more like CDO tranches. FTD baskets are more similar to CDO equity, albeit with less leverage in that the investor receives a premium in exchange for taking on first-loss risk. Even if fully funded, the default swap basket can be thought of as a leveraged investment in a basket of credits, with the non-recourse leverage being provided by the counterparty. The seller of protection in an FTD, however, is only exposed to the first credit event, and the principal loss is directly related to the recovery value of the defaulted asset. This aspect is unlike CDO equity, where the CDO structure is still active after the first loss and continues to be active until it matures, is called or amortizes. BASKET DEFAULT SWAPS ARE CORRELATION PRODUCTS
As in CDO tranches, correlation can have a dramatic effect on the pricing of basket default swaps. As an example, we consider a first-to-default swap on a basket of five credits, with each default swap trading at a premium of 50 bp. In the case of 0% correlation, a first-to-default swap would carry a premium of approximately 250 bp, the sum of the individual default swap premiums. In the case of 100% correlation, the first-to-default swap would have a premium of 50 bp, the maximum of the individual default swap premiums. A graph of the premiums on the first-, second-, third-, fourthand second-through-fifth-to-default swaps as a function of correlation is shown in Exhibit 2.
Please see additional important disclosures at the end of this report.
49
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 4
exhibit 2
Basket Default Swap Premium Depends on Correlation
Premium (bp)
First-to-default
300
Second-to-default Third-to-default
250
Fourth-to-default Second-through-Fifth
200 150 100 50 0 10%
20%
30%
40%
50% 60% Correlation
70%
80%
90%
Source: Morgan Stanley
FIRST-TO-DEFAULT BASKETS: INVESTOR MOTIVATION
Why would an investor either sell or buy first-to-default protection on a basket of credits? Say an investor has a positive view on a basket of five credits and that each credit pays a premium of 50 bp in the 5-year default swap market, as in the previous example. By entering into five individual default swaps, the investor could gain $2 million notional exposure to each credit and effectively receive 50 bp of premium per annum on $10 million notional for five years if no credit events occur. If one or more credit events occur, then there will be a principal loss in each default swap (equal to par minus the recovery value of the default asset). Alternatively, an investor can express a positive view on the same basket of credits by selling first-to-default protection on the basket with a notional amount of $10 million. From Exhibit 2, if we assume a correlation of 30%, the premium earned will be approximately 214 bp (86% of the portfolio spread); therefore, implementing this view via an FTD basket results in a higher-yielding investment for a given notional exposure. To get an equivalent yield from a single name, an investor would have to write protection on more than 4 times the notional. Therefore, an FTD is a much more levered way of expressing a positive view on credits, which is obviously a doubleedged sword. For other examples, refer to Chapter 74, where we show how an investor can get a return profile of BBs with a risk exposure similar to BBBs. Why would an investor buy protection on a basket of credits? It is effectively a way of shorting a basket of credits with less premium outlay than buying protection individually on each underlying credit. However, the swap gets terminated when the first credit event occurs, so the upside for the protection buyer is equal to par minus the recovery value of the defaulted credit only. This makes the basket a better hedge for default risk than spread risk.
50
Buying protection is also a way of hedging existing long exposure to credits, particularly for investors who want to reduce concentration risk but do not necessarily have an outright negative view on credits. For example, rather than buying protection individually on a portfolio of names, buying protection on the basket will be cheaper and more useful if the investor has a generally positive view on the credits but would like to reduce concentration risk. The buyer of protection is effectively short correlation. exhibit 3
Investor Motivation for Basket Trading
Basket
Protection Seller
Protection Buyer
FTD
Receive higher yield for same notional by taking more levered exposure to basket
Avoid idiosyncratic risk on a basket
Express credit views in levered form
Buy cheaper protection vs. individual swaps on all underlying names, but retain tail risk
Position for rising correlation (for example as a result of mergers)
Express views of falling correlation between individual names or industries
Express positive views on a basket, while avoiding idiosyncratic risk, but taking tail risk
Buy cheaper protection against the tail risk vs. buying protection on all underlying names
Receive possibly higher yield than some of the low yielding names in the underlying basket
Manage risk for extreme scenarios
Position for falling correlation (for example expressing views that expected mergers won’t materialize)
Position for rising correlation (for example as a result of mergers)
2nd-through-5th
Source: Morgan Stanley
MERGERS AND FTD BASKETS
The seller of protection on an FTD basket is long correlation, i.e., rising correlation among the names in the basket would result in a positive mark-to-market. This is also evident from Exhibit 2. This aspect of an FTD basket makes it an effective instrument for positioning for mergers and acquisitions activity among a set of names (assuming no replacement language, which we will elaborate on shortly). For example, if an investor expects M&A activity in a sector, he/she can write protection on an FTD basket, including names that are likely to be involved. If there is a merger, the correlation between the merged names will effectively go to 100%, resulting in an overall rise in basket correlation, benefiting the FTD protection seller. From another perspective, the FTD protection writer has to worry about only 4 credits after a merger, as opposed to 5 credits previously (assuming that the basket initially included 5 credits).
Please see additional important disclosures at the end of this report.
51
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 4
Replacement Language in FTD Baskets What happens if two of the names in a basket merge? The mechanics behind merger treatment in baskets are important to understand, as there are two forms that trade in the market (see Exhibit 4). In the simplest form (i.e., without replacement language), a five-name, first-to-default basket where two reference entities merge effectively becomes a four-name basket once the merger closes. exhibit 4
Millions 25
Replacement Language Matters – Impact of a Hypothetical Merger Between Credits 1 & 2
20 15 10 1
5
2
0
3
4
No Replacement
Replacement
Millions 25
Millions 25
20
20
15 10 5 0
6
15 1 & 2
5
Credits
2
10 3
4 Credits
5
5 0
1
3
4
5
Credits
Note: Assumes $10 million notional of first-to-default exposure. Source: Morgan Stanley
However, the “contractual” notional amount of the merged entity in the basket doubles to reflect that it was once two reference entities. Yet, this change in exposure size of the merged entity does not matter to the seller or buyer of first-to-default protection. A credit event in any of the four names (including the merged entity) would still trigger the contract. A credit event experienced by the merged entity, however, has a larger notional impact. Therefore, it would simultaneously trigger both a first-to-default and a second-to-default contract. Replacement Language – Operationally Puzzling The impact of a lack of replacement language seems intuitive to us, given either the correlation or leverage reduction analogies we described earlier. The seller of first-todefault protection sees a reduction in risk when there is a merger, so the seller of second-to-default protection must see an increase in risk, for which he or she must be compensated. Market clearing levels should take this into consideration, if the market is efficient about this. In fact, we would argue that the merger option is one factor that keeps implied correlation in small baskets higher than in large baskets.
52
First-to-default baskets trade with “replacement” language in the marketplace, as well (see Exhibit 3). Here, when a merger occurs, instead of dropping an entity from the basket, the buyer and seller of protection must mutually agree on a replacement credit, and often this replacement credit is one that trades at a similar spread level. Clearly, there is significant correlation risk in this exercise, and the buyer and seller of protection have opposite correlation views. Furthermore, the operational aspects of going through this process for each basket that experiences a merger can be tedious. We have much more experience with baskets that trade without replacement language, which, again, seems more intuitive to us. We thus recommend that both buyers and sellers of first-to-default protection avoid replacement language. The Merger Impact – What Is It Worth? In order to better understand the value of the merger options, we examine the pricing impact across credits with a variety of spreads and correlations. Generally, the merger impact becomes more valuable for riskier (e.g., higher spread) credits, as well as for baskets where implied correlation (before the merger announcement) is low. For example, the expected spread move in a basket with an average premium of 50 bp, at a correlation of 40% (which is typical for diversified industrial first-to-default baskets), would be 32 bp (see Exhibit 5). At 150 bp of average premium, it would be worth 80 bp. This assumes no movement in spread of the merged entity or other credits in the basket. exhibit 5
Correlation 20%
What Is the Impact of a Merger Worth? Average Spread of Credits in Basket (bp) 20 50 100 18 41 78
150 111
40%
15
32
57
80
60%
11
22
39
53
80%
6
13
22
29
Source: Morgan Stanley
To the extent that a merger drives the merged entity’s spread wider, the impact of the merger would be more muted than indicated in the table. As an example, for a basket with average spread of 50 bp, trading at 60% correlation, the gain on the merger of two entities (roughly 22 bp of first-to-default premium) would be fully offset by a 33 bp (66%) widening in spread for the merged entity. SECOND-THROUGH-FIFTH DEFAULT BASKETS: ANALOGOUS TO SENIOR CDO TRANCHES
A second-through-fifth default basket (assuming five credits) is similar to a senior CDO tranche, and any credit events after the first default trigger the contract. Unlike FTD baskets, the swap remains in effect until maturity or until all the covered credits default. The risk/return profile is similar to that of a CDO senior tranche in that a protection seller is protected from idiosyncratic risk, but exposed to severe loss scenarios where multiple credits default (also known as “tail risk”). However, a key difference is that the return profile of the protection seller in a second-through-fifth default basket is independent of the severity of the first loss.
Please see additional important disclosures at the end of this report.
53
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 4
SECOND-THROUGH-FIFTH-TO-DEFAULT BASKETS: INVESTOR MOTIVATION
Why would an investor sell or buy protection on a second-through-fifth default basis? As with senior versus equity in a CDO, selling protection in a second-through-fifth default basket is potentially a safer or less volatile way to gain long exposure to a basket of credits. In exchange for a smaller premium, the protection seller has direct exposure to any defaults after the first credit event, but is protected from the first credit event. An investor with a positive view on the underlying credits can sell protection in this swap and potentially earn more premium than selling protection on some of the individual names. In contrast to first-to-default baskets, the seller of protection is short correlation (i.e., an increase in correlation among the underlying names would be bad). Why would an investor buy protection on a second-through-fifth default basket? In general, it can be a significantly cheaper way of hedging existing long credit exposure when the investor is willing to take the risk of some losses on the first credit event, but wants to hedge against more severe default scenarios. FUNDED BASKETS
Most basket default swaps are structured and executed in “unfunded” form and are not assigned credit ratings. However, it is possible to combine an FTD basket with a deposit to create a funded product. In that case, the protection writer would buy a note with spread over Libor coming from an FTD position. In case of a default in the underlying portfolio, the note holder would incur a loss of principal on the note, reflecting the payment of par minus recovery on the FTD basket. EXAMPLE: EX-HIVOL FIRST-TO-DEFAULT BASKET
Among the more common types of first-to-default basket transactions we have witnessed in the marketplace are those involving high-quality credits from relatively unrelated industry groups. In such baskets, first-to-default protection sellers receive significantly higher premiums than the average for the basket because the low correlation among the credits effectively raises the probability that at least one credit will experience a credit event. In Exhibit 6, we show the Ex-HiVol basket as an example, with five investment grade credits and individual default swap premiums averaging 37 bp. Investors confident that none of the credits will experience a credit event over a five-year term can implement this view by selling first-to-default protection on the basket. In exchange for taking on the risk that any one of the credits experiences a credit event, the investor receives a premium of 153 bp. This premium is 4.1 times the average of the five names and 84% of the total, assuming correlation of 38%. The quoted prices can assume that the FTD protection seller/buyer would also buy/sell a delta-neutral amount of protection on the underlying credits. However, if the investor is not interested in entering into a delta-neutral transaction, the bid/ask would be wider, reflecting the transaction cost that the dealer has to incur for delta hedging.
54
Ex HiVol Benchmark First-to-Default Basket
exhibit 6 Names DOW
CDS Level (bp) 22
SRAC
90
DUK
25
GECC
21
VZ Total
Bid/Offer (bp) 153/164
Bid/Offer (% Total) 84/90
Imp Corr. Bid/ Offer (%) 38/25
25 183
Source: Morgan Stanley
Clearly, the correlation assumption is an important input in valuation of FTD baskets. In the next chapter, we will spend considerable time on developing intuition for correlation. As we mentioned earlier, sensitivities of baskets to spread changes, correlation changes, time decay, etc. are intuitively similar to those of CDO tranches, which we discuss in Chapter 6.
Please see additional important disclosures at the end of this report.
55
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 5
What Is Correlation? INTRODUCTION
Before we address what correlation is, it is worth a few words to address why correlation is of interest to credit investors. Portfolio structured credit products are generally based on portfolios of assets whose performance is interrelated. For example, default rates are economically cyclical and entire sectors can suffer from downturns, resulting in higher-than-expected default experience. These baskets of correlated assets behave differently from identical baskets of independent assets in several significant ways. Markets have long recognized these differences and prices have reflected the differing risks. Correlation is the missing link necessary to go from today’s standard derivative pricing models to the prices we observe in credit markets, or vice versa. WHAT EXACTLY DO WE MEAN BY CORRELATION?
We start with a definition from the American Heritage Dictionary: “Correlation: n. A causal, complementary, parallel, or reciprocal relationship, especially a structural, functional, or qualitative correspondence between two comparable entities: a correlation between drug abuse and crime. {Statistics} The simultaneous change in value of two numerically valued random variables: the positive correlation between cigarette smoking and the incidence of lung cancer; the negative correlation between age and normal vision.” CORRELATION IN FINANCIAL MARKETS
In financial markets, correlation is best understood using the latter statistical definition. Most often, correlation in financial markets describes the relationship of the price changes (or returns) of two different assets. Consider the classic CAPM model, which implicitly uses the idea of correlation between the return of a given asset and the overall market in the definition of beta: Beta
= Cov(Ri,Rm) / σ2(Rm) = ρ(Ri,Rm)* σ (Ri) / σ(Rm)
where Ri and Rm are the return on asset i and the market, respectively Cov is covariance, ρ is correlation, and σ is standard deviation
This is the type of correlation many investors and asset allocators are most accustomed to, and it has long been discussed in the financial literature. Another example of this same type of correlation is in the world of equity index options where we can examine the correlation between securities, rather than the
56
correlation between a security and the market. Consider the market for options on the Dow Jones Industrial Average and all of the components. Traded option markets tell us that the weighted average implied volatility for the portfolio was roughly 21% on March 23, 2005 while the observed implied volatility for options on the actual index was roughly 11%. The difference is explained by the correlation among the components. In Exhibit 1, we summarize the implied index volatility as calculated under several correlation assumptions, as well as the observed index volatility from the options market. If the credits are 100% correlated then there are no diversification benefits and the volatility of the portfolio is simply the weighted average of the components, which is 21.3%. If the components are independent of one another, there are significant diversification benefits, with an implied index volatility of 4.3%. The reality is somewhere in between, and on this particular day, the correlation implied by market pricing was approximately 24.5%. exhibit 1
Portfolio Volatility Affected by Correlation
Correlation 0.0%
Index Volatility 4.3%
50.0%
15.3%
100.0%
21.3%
24.5% (Index Option Implied)
11.0%
Source: Morgan Stanley, Bloomberg
In order to simplify the analysis above, we have used one common correlation variable for each pair of components and varied that number. In reality, this assumption of a single correlation between these assets is probably inadequate and the market pricing likely reflects the varying levels of correlation among the assets in the portfolio. While historical correlations among the returns of the index components are observable (because we have active equity markets for all the components), these are backwardlooking estimates, like realized volatility. CORRELATION IN STRUCTURED CREDIT MARKETS
Correlation in structured credit markets has a very different meaning than in most financial markets. We highlight the key differences below: 1 First, the correlation quoted in structured credit markets is generally implied from the market price of a tranche, under a certain set of assumptions using a given model. Therefore it is usually an output, rather than an input, and it is generally forwardlooking rather than historical. This may not have been the case in the early days of the market or in some of today’s bespoke structures, but standardization in both models and contract terms has occurred almost organically, allowing market participants to quote correlations to communicate price, given all the other market variables. Interpreting an implied correlation generated by one of today’s standard copula models 1
This discussion focuses on the correlation measure used in standard copula implementations rather than rating agency approaches. For a discussion of rating agency approaches, please see Chapters 1 and 2.
Please see additional important disclosures at the end of this report.
57
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 5
requires market data or assumptions regarding all of the variables listed in the Valuation section on Page 61. Default Correlation Not Spread Correlation Second, and probably most important, the correlation we are actually measuring is not the correlation of the change in prices (or spreads) but rather it is most closely related to the correlation of the timing of default events within the portfolio. In contrast to equity derivative models, structured credit models generate values based on default probabilities derived from today’s market prices of the underlying instruments rather than the expected value based on the future market prices of the underlying instruments. This is a fundamental difference in the way credit derivatives are valued as compared to derivatives based on other asset classes. The current generation of structured credit models rely on the distribution of aggregate default losses for the portfolio to derive the tranche price. Aggregate losses for the portfolio are generated by combining the likelihood of default for each credit with a measure of how correlated the default events are. The expected timing of default events for the individual credits are directly related to today’s spreads for the portfolio components (but do not reflect the potential movement in spreads in the future). Aggregate Correlation – If You Can Call It That Third, the quoted correlation is a single number that represents the relationship between all the pairs of names in a given portfolio. Correlation is generally a concept that measures the relationship between any two assets. In structured credit markets, the same number is used to represent the relationship between every pair in the portfolio (as we have done in our Dow Jones Industrial Average example). When we say the price of the tranche has an implied correlation of 30%, we are essentially saying that the correlation for each distinct credit pair is implied by the market price to be 30%. While this definition does not allow us to vary correlation for each pair, it does provide a measure of the aggregate correlation in the portfolio. As with the equity index example above, there are likely differing levels of correlation for the various pairs of names, but deriving correlations for all the pairs can be an exercise in futility. Unlike the equity index example, we cannot actually observe the historical default correlation for two given credits (if a name had a correlated default to observe, would it be in the portfolio?) so approximations of the pair-wise correlations add the risk of significant estimation errors. We can observe spread correlations, assuming spread markets are liquid, but these are imperfect proxies for default correlations (which we discuss below). Additionally, the sheer number of estimates required makes the process an unmanageable exercise fairly quickly. Consider a typical 100-name portfolio, which would require 4,950 distinct correlation estimates. For all but the most junior tranches, the aggregate correlation of a portfolio is a more important driver of pricing than the correlation between any two credits. For the most subordinate tranches, the correlation between a single pair of credits can be an important driver, as well, given their exposure to idiosyncratic risk. For example, if the two widest trading names in a basket announced a merger for which the financing plan indicated increased leverage (and by implication, higher levels of default risk) for the joint entity, this would be doubly bad for the most junior tranches, while having only a
58
marginal impact on the most senior tranches because of the idiosyncratic nature of the risk. Default propensity increased only for the two names and not for the broader universe, and correlation jumped between only these two entities, but their correlation to the rest of the portfolio likely dropped. Consider how muted the benefit would be to the same tranche if two of the most high-quality credits announced a credit-enhancing merger. Situations like this help, in part, explain the difference in implied correlations for actively traded tranches of the same portfolio. OBSERVING DEFAULT CORRELATION
We like to say that implied correlation in structured credit is similar to implied volatility in equity options markets. It provides a way to communicate price, given all the other inputs, which are generally observable. Under this framework, observing different implied correlations for different tranches is like observing different implied volatilities in options markets: it measures the supply and demand dynamics in the market and, if persistent, can point to weaknesses in the assumptions or the model we are using. There is one critical difference, however, between implied volatility and implied correlation. If we want to test implied volatility against realized volatility in markets, we can do so. We can observe the returns in the market during the period and compare realized return volatility against that implied by option prices. This type of comparison is not possible in structured credit markets, largely because defaults are generally low probability events and are hence infrequent, particularly in investment grade markets. What we really want to observe is the correlation between two events, each of which is itself fairly unlikely. Users in today’s market have turned to observable data to get a sense for default correlation as a proxy for a variable that is difficult to observe. This can be an exercise fraught with peril. To illustrate the point, we again turn to the equity markets. We examine two measures of correlation for the current components of the Dow Jones Industrial Average from January 1995 through March 2005. The first metric, which we will call “price correlation” is the average correlation of the daily percentage returns for each pair of index components. This measure is consistent with the correlation that is most frequently encountered in financial markets. The second metric, on the other hand, is designed to be similar to the concept of the time to default correlation we use in today’s structured credit models. It is defined as the average of the “timing correlations” for the components, where “timing correlation” is defined as the correlation of the number of “extreme return” days in a given month for each pair of index components. We define “extreme return” as a daily return greater than +/-5%. The idea is for this metric to capture the correlation of extreme daily moves only, assuming a month is a sufficient period to capture any lag among the components. For the full period in question, we find that the average return correlation for the DJIA is roughly 16%, while the timing correlation is roughly 34%, or more than double the price correlation. We find similar (albeit flipped) results when we examine correlations in spread markets: the realized spread (price) correlation for the components of DJ CDX NA IG 3 was 56% for the year ending March 29, 2005 while the implied (timing)
Please see additional important disclosures at the end of this report.
59
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 5
correlation in the 0-3% tranche on that date was 19%. The implication is that day-today equity price changes are only slightly correlated, while extreme events in equities are much more correlated. For credit spreads, the opposite is true: daily price changes are highly correlated, but default events are less so. The point here is really that when we talk about the correlation of extreme events like default, we need to be careful extrapolating from correlations that capture both large and small changes in valuation, such as the correlations of the prices of equities, assets or spreads. THE HIDDEN MEANING OF DEFAULT CORRELATION IN CREDIT MARKETS
After such a long-winded discussion of exactly what “correlation” is in structured credit markets, it is worthwhile to highlight what correlation does to the implied behavior of structured credit portfolios. Some work we have done in the portfolio optimization space, in which we focus on portfolio size, helps highlight the impact of correlation on credit portfolio performance. In Exhibit 2, we show the volatility of expected losses of similar investment grade portfolios with different numbers of independent credits. As we increase portfolio size from 50 names to 1,000 (think of 1,000 as the investable universe), the uncertainty of expected losses of the portfolio declines to some minimum level (0.4% over five years for the largest portfolio). We then perform the same analysis for the same portfolio and superimpose 20% default correlation on the portfolio. The results in Exhibit 2 show that the absolute level of uncertainty declines much less than in the independent case. This points to the first impact of correlation on the credit portfolios, namely, that the absolute level of risk is greater than for an independent portfolio (this may sound obvious). Alternatively, the diversification impact of adding names to the portfolio is much less important for correlated names. exhibit 2
Correlation Affects the Relationship Between Volatility and Portfolio Size (Expected Losses = 3.0%)
Standard Deviation of Expected Losses 4.0% 3.5% 3.0% 2.5%
Independent 2.0%
20% Correlated
1.5% 1.0% 0.5%
950
1000
900
850
800
750
700
650
600
550
500
450
400
350
300
250
200
150
50
100
0.0% Portfolio Size
Source: Morgan Stanley
The real insight comes from examining these results against real world experience. We summarize the coefficients of variation (standard deviation divided by mean) of the simulation results in Exhibit 3 alongside the same for Moody’s five-year cumulative default statistics for the 1970-1999 period. The most valid comparisons are the 1,000name portfolios against the Moody’s results, because the 50-name portfolios have much more idiosyncratic risk. Most notable is that the volatility of expected losses for
60
large portfolios is much greater for correlated portfolios and this is broadly consistent with actual default experience we see from Moody’s. This is the hidden impact of correlation: to make the aggregate default experience inside our models more similar to what we actually experience, given the cyclical (correlated) and volatile nature of credit risk. Volatility – In the Models and in the Real World
exhibit 3
50 50 1000 1000 Credits 0% Credits 0% Credits 20% Credits 20% Correlation Correlation Correlation Correlation Coefficient of Variance
0.6x
0.1x
1.1x
1.0x
Moody's A
Moody's Baa
Moody's A & Baa
1.5x
0.7x
0.9x
Source: Morgan Stanley
VALUATION IN STRUCTURED CREDIT MARKETS
Valuation for the structured credit market has developed along several tracks. The most recent introduction is in the synthetic space and has been based largely on a derivative pricing framework not dissimilar to that used to price single-name credit derivatives. As with single-name credit derivatives, pricing for tranches or N-th to default baskets is a function of the single-name par spread levels, the spread at which the contract is struck, the risk-free interest rate and a recovery assumption, but we need to also add the following variables for tranches: attachment and detachment points (either in losses or in number of defaults) and the correlation of default events. As with the single-name models, the spreads are converted into risk-neutral default probabilities, which is effectively the same as generating a loss distribution for the single-name credit. We can think of the single-name default probabilities as being represented by a very simple Bernoulli probability distribution, where default for a given single name is distributed in the following way: 1 if default with probability p 0 if no default with probability 1-p Therefore, losses are distributed as follows, using a fixed recovery assumption: (1-recovery rate)*notional amount if default with probability p 0 if no default with probability 1-p The key to valuing tranches of a given portfolio is the ability to generate a distribution for the number of defaults in the entire portfolio over a given time horizon. This, in turn, gives us the ability to calculate the probability that losses in the portfolio exceed given thresholds (like the attachment and detachment points of tranches). Once we have these probabilities of various levels of default for the portfolio, we can generate expected losses, cashflows and value tranches accordingly. GETTING TO A PORTFOLIO LOSS DISTRIBUTION
In order to explore why copula models for structured credit products were developed, we begin with a simplified example and make it more complex to better fit reality. If
Please see additional important disclosures at the end of this report.
61
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 5
we assume the default probability for each credit in a given portfolio (say, with 100 names) is the same and all the credits are independent of one another, generating the distribution of portfolio losses from the single name default probabilities is a simple matter. Going back to our statistics classes, if we are willing to assume that the singlename loss distributions look like Bernoulli trials, then the sum of 100 Bernoulli variables with the same probability of default has a binomial distribution (check with your local statistics expert if you don’t believe us). We then add the complication that the probabilities of default vary by credit and the problem becomes more complex. We no longer have a clean and simple closed form solution, but we can still solve the problem using simulation (or numerical approximations). The simulation process is fairly straightforward. We simulate 100 independent random numbers, check whether the random number for each credit generates a default for that credit and add up the individual losses caused by each default to generate a loss for the entire portfolio. We then repeat the process many times to generate a distribution for portfolio losses like those graphically illustrated in Exhibit 4. exhibit 4
Correlation Creates Portfolio Loss Distributions with Fatter Tails
Probability Independent & Identical
60%
Independent Correlated
50% 40% 30% 20% 10%
20-22%
18-20%
16-18%
14-16%
12-14%
10-12%
8-10%
6-8%
4-6%
2-4%
0-2%
0%
Losses
Source: Morgan Stanley
In Exhibit 4, we show the aggregate loss distributions for portfolios in which all the credits are independent and have the same default probability (5.5%) and a portfolio for which all the credits are again independent and have different default probabilities (although with an average default probability of 5.5%). Both were generated using simulation with 10,000 trials. The portfolios underlying these results are similar to the investment grade CDX index products in that the average spread is 66 bp and, in the second case, the distribution of spreads is similar to the CDX portfolios. What should immediately jump out is the similarity of the distributions but also the fact that the distribution of losses in both cases is very concentrated, with more than 50% of the scenarios inside the 4-6% range. The third set of data in the exhibit reflects the idea that the individual credits are not independent but are positively correlated to one another. There a few key takeaways worth noting from the exhibit. First, the resulting portfolio loss distribution is much less concentrated with no single bucket accounting for more than 36% of the simulations. Additionally, the distribution has a much fatter tail. To illustrate the point,
62
consider the likelihood of aggregate portfolio losses exceeding 10%, which is less than 0.1% for both scenarios in which defaults are independent. This compares to roughly 8% for the correlated scenario. The probabilities from the independent simulations would imply that the spread on tranches with attachment points above 10% are nearly zero, so positive correlation is a way to explain the pricing we observe in the marketplace. It is also consistent with the idea that credit risk is driven by cyclical changes in the underlying economy. COPULA FUNCTIONS
In order to generate the results in the exhibit we applied a simulation-based approach, which reflected positive correlation among default events. In order to do this, we need some way to generate correlated random numbers from which to generate the default scenarios. This is where copula functions come into play. Copula functions are mathematical functions that allow us, in a simulation context, to transform independent random numbers into correlated random numbers. While copula functions can be derived for several statistical distributions, the most common implementation (in structured credit markets) involves combining many normal distributions into a correlated multivariate normal distribution (also referred to as the Multivariate Gaussian Copula). These correlated normal variables are then mapped into default times based on the spreads (and implied default probabilities) for the underlying credits. These correlated single-name default times can be mapped into losses for a contract with a given maturity, which can, in turn, be aggregated into portfolio losses for the same given maturity. In a simulation context, we repeat this process many times to generate the portfolio loss distribution in Exhibit 4. Why do we use the normal distribution? As with most other derivative models, the computation efficiency and theoretical attractiveness of the normal distribution were probably important drivers in selecting this as the basis for most structured credit pricing models. There are obviously alternatives to this, with the student T copula function being the most notable candidate. CORRELATION AS A RELATIVE VALUE METRIC
The language of implied correlation has evolved to provide a useful relative value metric in the form of the often quoted base correlation. The first time we heard the term “implied correlation,” it was used to refer to what we now call “compound correlation” – that is, the correlation implied by the price of a tranche, given all the other market inputs. The problem with compound correlation is that it fails to give meaningful results in some of the mezzanine parts of the capital structure. In Exhibit 5, we summarize the implied compound and base correlation for one of the first liquid traded index products, Dow Jones TRAC-XSM Series 2. We find it notable that the implied correlation for the 3-7% tranche is so much different from the other observations.
Please see additional important disclosures at the end of this report.
63
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 5
Implied Compound and Base Correlation
exhibit 5
Compound Correlation
Base Correlation
47%
21%
21%
435
3%
25%
138
158
17%
28%
56
66
21%
34%
Index
Bid 68
Ask 69
0-3%*
43%
3-7%
395
7-10% 10-15%
*Point upfront plus 500 bp running. Source: Morgan Stanley
In Exhibit 6, we show the spreads implied by compound correlation levels ranging from 10-90%. We find that there are actually two possible solutions for the 3-7% with a spread of 415 bp. This complicates the use of compound correlation as a relative value measure. Additionally, the idea of using compound correlation as a relative value metric is even further complicated by the fact that rising correlation is good for some tranches and bad for others, as the differing slope of the lines in Exhibit 6 illustrates. exhibit 6
Correlation Sensitivity Varies with Seniority
bp 900 800
2-5%
3-7%
7-10%
10-15%
15-30% 700 600 500 400 300 200 100 0 10%
20%
30%
40%
50%
60%
70%
80%
90%
Correlation
Source: Morgan Stanley
Base correlation, which uses the implied correlations from a series of first loss tranches with detachment points equal to the detachment points of the actual tranches, does not suffer from either of the two issues above. 2 It does not have tranches for which there are multiple solutions because all the correlations are implied from first loss tranches. We provide an example of one implementation of a base correlation approach in Exhibit 7. Base correlation provides a good relative value benchmark because it is fairly uniform in its meaning (i.e., rising base correlation means tighter spreads on that tranche, all else being equal), and it provides a unique price for a given base correlation. Additionally, it provides price sensitivities that are more in line with what we have observed in the marketplace when compared to compound correlation, which can imply unreasonable price sensitivity values for very high or very low implied correlation values. 2
64
Please see Chapters 13 and 22.
exhibit 7
Benchmark Tranche 3-7% 7-10% 10-15%
First Loss Tranche 0-3%
Tranche Equivalence – Long/Short Combinations Replicating Portfolio First Loss Imp. Tranche Correlation Position Short 21.2%
First Loss Delta 12.2x
0-7%
Long
27.5%
10.1x
0-7%
Short
27.5%
10.1x
0-10%
Long
31.6%
8.3x
0-10%
Short
31.6%
8.3x
Implied Benchmark Tranche Delta 8.5x 3.9x 1.9x
Source: Morgan Stanley
WHAT IS CORRELATION MISSING?
In structured credit markets, it has never been the case (at least to our knowledge) that all the tranches of a given index/portfolio traded to the same correlation. There are several likely explanations for this (aside from liquidity), but most relate to assumptions in today’s standard models. The use of a single correlation number fails to capture the fact that there are stronger relationships between specific companies and even entire sectors. These subtleties are more important for the more subordinate parts of capital structures and can move prices in ways that force our single implied correlation metric to move away from the correlation in other parts of the capital structure. Today’s copula models are risk-neutral in nature and therefore assume that investors are indifferent when faced with a choice between risky assets that can be hedged and risk-free assets. While single-name CDS are clearly a traded commodity, pricing in this market is far from continuous and investors in subordinate portions of capital structures have levered exposure to the risk that default swap prices in the underlying portfolio jump. Lower correlation for more subordinate portions of the capital structure can reflect the increasingly levered exposure to (and subsequent risk premium charged for) jumps in single-name prices or exposure to idiosyncratic risk. Another factor in the relative pricing of tranches is the willingness of rating agencies to assign a rating to a particular tranche. The most severe dislocations in the pricing (and the largest implied correlation skew) in the market for liquid tranches occur at points between what are typically unrated portions of the capital structure and what are typically the most subordinate rated portion of the capital structure. In this case, the difference of correlation skew simply represents the different supply and demand dynamics for instruments with differing risk levels. The rating may add an institutionalized component, exacerbating these supply/demand dynamics and increasing the skew in correlation. Finally, most of today’s standard models use the normal distribution (or the Student T) to generate correlated defaults. It is very possible that the assumed distribution introduces error into the models simply because the distribution is not the best approximation of the real world.
Please see additional important disclosures at the end of this report.
65
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 6
Understanding Tranche Sensitivity Having introduced the market standard correlation model in Chapter 5, we can now discuss how to calculate relative sensitivity measures based on this model, in the typical language of option greeks. The inclusion of tranches in portfolios has enabled investors to express credit views with very different risk and return profiles. For example, it is now possible to separate default risk from spread risk, and to take credit positions that become long when spreads are falling and short when spreads are rising. This precision, necessary for the implementation of such sophisticated strategies, emanates directly from the differences in sensitivities of tranches. exhibit 1
Summary of Greeks
Greek Delta (į)
What does it measure? Tranche price sensitivity to changes in underlying portfolio spreads, measured as a ratio of tranche PV01 to index PV01 (PV10% is also used sometimes)
M-Gamma (m-Ȗ)
Tranche price sensitivity of a delta-neutral position to parallel shifts in spreads of underlying names. It represents a form of convexity (M = Market)
I-Gamma (i-Ȗ)
Tranche price sensitivity of a delta-neutral position to jump-to-default risk or changes in spread distribution of the underlying portfolio. It represents a form of convexity to moves in a single credit while all others remain constant (I = Idiosyncratic risk)
Rho (ȡ)
Change in tranche value due to changes in default correlation
Theta (ș)
Change in tranche value due to the passage of time
Source: Morgan Stanley
While we can borrow the basic concepts from the world of equity derivatives, where option sensitivities or “greeks” are well-documented and understood, the greeks for tranches of credit portfolios have a distinct flavor. Exhibit 1 shows the basic definitions of various greeks for structured credit. In this chapter, we start by exploring how tranches can be viewed as options on defaults and then delve deeper into the greeks. We have deliberately taken an intuitive approach in our explanations, as opposed to a mathematical one. Furthermore, while we focus largely on standardized tranches for simplicity, the arguments we make in this chapter hold for similar synthetic CDO tranches.
66
TRANCHES AS OPTIONS ON DEFAULT
One way of thinking about tranches is to view them as options on default losses on the underlying portfolio/index. This perspective helps us develop a feel for the sensitivity of tranches to various variables, such as time decay, spread changes, defaults exposure, etc. We can easily estimate expected losses for an index from its spread level, which serves as a benchmark for the tranches of the index. For example, a spread level of 50 bp on the investment grade Dow Jones CDX fiveyear index implies losses of roughly 2.5% for five years on the portfolio (ignoring discounting and lost premiums in case of defaults), in a risk-neutral sense. Therefore, the 0-3% tranche is expected to lose some notional due to defaults and can be viewed as in-the-money. By the same argument, more senior tranches are out-of-the-money. Interestingly, as the index spread widens, expected losses would increase and senior tranches could expect some notional loss, depending on the increase in spreads. Consequently, the subordination level that separates in-the-money and out-of-themoney tranches could change. For example, let us assume that the index widens to 100 bp, translating to roughly 5% losses. Now the mezzanine tranche, 3-7%, is also expected to lose some notional due to defaults. In other words, the tranche is now inthe-money. This implies that while the mezz tranche behaves more like senior tranches at low spread levels, it would act more like an equity tranche as spreads widen significantly. Different index tranches have different capital structures, implying that the subordination level that separates in-the-money and out-of-the-money tranches may be different. For example, the 0-10% and 10-15% tranches in High Yield DJ CDX are inthe-money (i.e., the index level implies that defaults would penetrate these tranches), implying that they behave more like an equity tranche, while the next tranche, 15-25%, behaves more like the investment grade mezzanine tranche. Similarly, the mezzanine investment grade 3-7% five-year tranche is out-of-the-money, while it is in-the-money for the 10-year index, again implying markedly distinct sensitivities. For the same reason, as time passes and spreads roll down the credit curve, tranches could move from in-the-money to out-of-the-money and their greeks would also change. Again, the intuition about how in-the-money and out-of-the-money tranches behave makes the process of projecting changes in tranche sensitivity more approachable. SENSITIVITY TO SPREAD CHANGES OR “DELTA”
As the spread of the underlying index changes, the impact on different tranches varies. Overall, it is consistent with the index directionally, i.e., as spreads widen, a short protection position in any of the tranches would lose money. Conversely, as spreads tighten, a short protection position would see a positive mark-to-market.
Please see additional important disclosures at the end of this report.
67
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 6
For small spread movements, one can estimate the impact using tranche delta, which is the ratio of change in the tranche value to change in the index for a given spread change. While it is usually calculated as the ratio of PV01s, some participants also use PV 10%. (PV01 is the change in the mark-to-market for a 1 bp move in each of the underlying credits in the portfolio. PV10% is the change in mark-to-market for a 10% increase.) Tranches with higher deltas would move more than tranches with lower deltas. Furthermore, tranches with deltas less than 1x would move less than the index, while tranches with deltas higher than 1x would move more than the index. However, for bigger moves in spreads, the delta-based calculation is only approximate, as the impact of tranche convexity becomes more meaningful, which we will discuss later in this chapter. exhibit 2
Spread Sensitivity
P&L(%) 25% 20%
0-3%
3-7%
7-10%
10-15%
15-30%
0-100%
15% 10% 5% 0% -5% -10% -15% -20% -25% 50%
75%
100%
125% 150% Spread Change Factor
175%
200%
Source: Morgan Stanley
Broadly speaking, junior tranches have higher deltas than senior tranches, assuming both are quoted on a running premium basis. Given that junior tranches take losses before senior tranches, wider spreads make them proportionally more likely to be affected by the increased likelihood of defaults implied by wider spreads. In other words, the “default PV” (PV of expected default losses) rises faster than the index for junior tranches, but slower than the index for senior tranches, bringing the average default PV for tranches inline with the index default PV. Additionally, as expected losses rise, the likelihood of lost premiums due to lower notional outstanding (as a result of defaults) also rises, which further accentuates the negative impact of rising spreads. Impact of Upfront Payments However, the presence of upfront payments for tranches lowers their deltas, compared to the same tranches without an upfront payment. Currently, the equity tranche for the investment grade DJ CDX index and the first two tranches for the high yield DJ CDX index trade with upfront payments (for more details, see Chapter 3). Why does an upfront payment lower the delta of a tranche? The answer lies in understanding how rising spreads affect the present value of expected premiums (premium PV) and default PV. An upfront payment with no running coupon implies that as spreads widen, the tranche value is affected only by higher expected defaults,
68
and not lower expected premiums, since the protection seller has already collected all of the premium. On the other hand, an all-running contract will have both negative impacts – higher expected defaults and lower expected premiums due to more defaults making it more sensitive. “I-GAMMA” OR SENSITIVITY TO SPREAD DISTRIBUTION CHANGES
While the impact of overall spread changes on tranches is more or less obvious, the effect of changes in distribution of the underlying spreads, especially when the overall portfolio average remains unchanged, is subtle and worth exploring further, as such moves are more common than big swings in the indices. Tight trading names moving somewhat wider generally impact senior tranches, while wide or even average credits moving significantly wider impact junior mezzanine and first-loss tranches, depending on the size of the move. A good example of actual occurrence was the reshaping of the risk profile of investment grade in October 2004 due to stress in the insurance sector, especially with Marsh & McLennan (for details, see Chapter 44). Thus, the shape of the risk distribution within the portfolio influences the pricing of tranches (see Exhibit 3). It should be fairly intuitive to recognize that the length and thickness of the right tail influences the pricing of subordinate tranches, meaning that the bigger the tail, the riskier the equity and mezzanine tranches. Similarly, the shape of the middle to left side of the distribution should influence the more-senior notes. As credit risk increases (shifts from left to right), senior tranches should become riskier. exhibit 3
Insurance Moves to the Right, Impacting Both Senior and Subordinate Tranches
Number of Credits 30 25 20 15 15-30% 7-15% 3-7% 0-3%
10 5
190 - 200
180 - 190
170 - 180
160 - 170
150 - 160
140 - 150
130 - 140
120 - 130
110 - 120
100 - 110
80 - 90
90 - 100
70 - 80
60 - 70
50 - 60
40 - 50
30 - 40
20 - 30
0 - 10
10 - 20
0
Spread
Source: Morgan Stanley
Two credits in the insurance sector served as useful examples. The move in AIG’s default swap premium (from 18 bp to 34 bp) increased risk in 15-30% type tranches, while reducing risk in super-seniors (30-100%). On the other hand, Marsh & McLennan (which widened from 30 bp to north of 250 bp) was clearly a right tail event, shifting risk from 15-30% type tranches to 0-3% (hypothetically, since MMC was not a part of the index). If we take this analysis one step forward, we can see the hypothetical pricing impact on tranches (see Exhibit 4).
Please see additional important disclosures at the end of this report.
69
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 6
exhibit 4
0.05%
Tranche Pricing Impact: Two Opposite Examples
0.00%
-0.05%
AIG+16bps -0.10%
MMC+240bps
-0.15%
-0.20% 0-3%
3-7%
7-10%
10-15%
15-30%
Source: Morgan Stanley
DELTA MIGRATION
In our past publications we have discussed the issue of instability of tranche deltas several times. Broadly speaking, tranche deltas change with movements in index spreads, passage of time, and changes in correlation. Consequently, trades that are delta neutral at inception can easily become delta positive or negative, depending on these factors. Understanding these sensitivities can help us fine-tune investment strategies to achieve desired delta under different scenarios. For example, it is possible to construct trades that become delta positive when spreads tighten and delta negative when spreads widen, resulting in a more desirable payoff. Similarly, suppose an investor believes that the spreads are currently too tight and will widen in the near future but start tightening down the road, say next year. It is possible to implement a tranche combination that starts out delta negative but becomes more and more delta positive as time passes.
70
exhibit 5
Delta Sensitivity– 5 Year IG DJ CDX
Delta (x) 30
0-3% 7-10% 15-30%
25
3-7% 10-15% 0-100%
20 15 10 5 0 50%
75%
100%
125% 150% Spread Change Factor
175%
200%
Source: Morgan Stanley
Let us first analyze how index level affects tranche deltas. Exhibit 5 shows how deltas change with spread movements. The intuitive way to think about delta sensitivity to spread changes is that the deltas for in-the-money tranches decrease with spread increases, while deltas for out-of-the-money tranches increase with spread increases. Additionally, one has to bear in mind that more senior tranches start becoming in-themoney after a certain level of spread increase. This point is easier to see in the delta sensitivity pattern for 10-year investment grade DJ CDX tranches, as shown in Exhibit 6. Another point worth noting is that beyond a certain level of spread widening, the equity tranche delta declines rapidly. This is because expected losses far exceed the tranche notional, and the impact of any increase in expected losses is marginal. exhibit 6
Delta Sensitivity – 10 Year IG DJ CDX
Delta (x) 16 14
0-3%
3-7%
7-10%
10-15%
15-30%
0-100%
12 10 8 6 4 2 0 50%
75%
100% 125% 150% Spread Change Factor
175%
200%
Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
71
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 6
How do tranche deltas change as time elapses? Exhibit 7 shows how tranche deltas migrate as the index approaches maturity. For simplification, we have assumed that the credit curves for the underlying portfolio and tranche base correlations remain unchanged. Once again, it is more intuitive to think of the results in terms of tranches being options on defaults. Deltas for tranches that are in-the-money rise as time passes, while deltas for out-of-the-money tranches decline over time. Furthermore, since we have assumed that spreads roll down the curve, senior tranches get more and more outof-the-money over time. (Note that we have assumed that there are no defaults.) Another way to understand this delta migration pattern is to look at delta sensitivity to spread movements (Exhibit 5). As we roll down the curve index, spread declines resulting in rising delta for equity tranches and falling deltas for more senior tranches. exhibit 7
Delta Sensitivity Over Time – 5 Year IG DJ CDX
Delta (x) 40 35 30 25 20 15
0-3%
3-7%
7-10%
10-15%
15-30%
10 5 0 -5
Current
1 Year
2 Years
3 Years
4 Years
5 Years
Time Passed
Source: Morgan Stanley
Now that we understand the delta migration pattern over time, it is possible to design strategies with desired spread sensitivities over time as we alluded earlier. For example, by combining delta-neutral notional amounts of equity and senior tranches we can put together a trade that becomes delta positive as spreads ride down the curve, resulting in a positive mark-to-market. Similarly, we can design positions with a delta mismatch initially, so that the position becomes delta negative or positive for an expected spread move or passage of time (see Exhibit 8, for example).
72
exhibit 8
Delta Migration Over Time
Delta (x) 0.8 0.6 0.4 0.2 0.0 Current
1 Year
2 Years
3 Years
-0.2 0-3% Delta Neutral -0.4
0-3% Delta Negative
-0.6 Time Passed
Source: Morgan Stanley
“M-GAMMA” OR CONVEXITY
As we discussed earlier in this chapter, for wide spread moves, the relationship between tranche value and index value does not remain linear. This difference between linear approximation (using deltas) and the actual movement in market value is captured by a tranche’s convexity or gamma. We typically measure it as a ratio of PV100 to 100xPV01, i.e., the ratio of tranche mark-to-market for a 100 bp move in underlying spreads to 100 times PV01 of the tranche. exhibit 9
5 Yr IG – Tranche Convexity (Delta Neutral)
P&L (%) 0.6%
0-3%
3-7%
7-10%
10-15%
15-30%
0.4% 0.2% 0.0% -0.2% -0.4% -0.6% -0.8% 50%
75%
100% 125% 150% Spread Change Factor
175%
200%
Source: Morgan Stanley
The easiest way to observe the materiality of convexity is to plot the P&L of a deltaneutral transaction of tranches, as we have done for the 5-year DJ CDX tranches in Exhibit 9. As mentioned earlier, tranches that are in-the-money are positively convex, while out-of-the-money tranches are typically negatively convex (from the perspective of the protection seller). In other words, a delta neutral short protection position in fiveyear IG equity tranche would have a positive mark-to-market for large changes in spreads, while a delta neutral 7-10% tranche of the same index would have negative mark-to-market.
Please see additional important disclosures at the end of this report.
73
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 6
exhibit 10
10 Yr IG - Tranche Convexity (Delta Neutral)
P&L (%) 5% 4%
0-3%
3-7%
7-10%
10-15%
15-30% 3% 2% 1% 0% -1% -2% 50%
75%
100% 125% 150% Spread Change Factor
175%
200%
Source: Morgan Stanley
Exhibit 10 shows the convexity characteristics of 10-year IG tranches, where both 0-3% and 3-7% tranches are positively convex. To further illustrate this point, in Exhibit 11, we have shown the convexity attribute of the 3-7% 5-year and 10-year tranches. Clearly, the 3-7% tranche is negatively convex (and out-of-the-money, given current spread levels) for five-year DJ CDX, but it is positively convex (and in-themoney) for 10-year maturity. exhibit 11
Mezzanine Tranche Sensitivity
P&L (%) 0.6% 0.5%
3-7% 10 yr
0.4%
3-7% 5 yr
0.3% 0.2% 0.1% 0.0% -0.1% -0.2% -0.3% -0.4% 50%
75%
100% 125% 150% Spread Change Factor
175%
200%
Source: Morgan Stanley
While trading tranches of different underlying indices, for example 10-year tranches versus five-year tranches, convexity can be hard to measure, as the underlying indices are different and have convexity differences themselves. In addition, curve shape changes are also important to consider. For example, a steepening of the 5s-10s credit curve would affect 10-year tranches but not the five-year tranches.
74
JUMP-TO-DEFAULT SENSITIVITY
While higher default sensitivity for junior tranches and lower for senior tranches is intuitive, given their relative positions in the capital structure, comparing sensitivities across different maturities and indices provides valuable insight into how defaults affect different tranches. For example, one default shaves off 14% of the value from a five-year investment grade equity tranche, but only 7% from a 10-year tranche and 2% from a five-year high yield equity tranche. In Exhibit 12, we have summarized the average impact of one default on various tranches, assuming a 40% recovery for investment grade and a 35% recovery for high yield. Exhibit 13 shows the same impact on a delta neutral basis, i.e., assuming that we bought protection on the index to offset the tranche’s spread sensitivity. exhibit 12
Loss Due to 1 Default (% of Notional)
-14.0% -12.0% -10.0% -8.0% -6.0% -4.0% -2.0%
Index 5 Yr HY
25-35% 5 Yr HY
35-100% 5 Yr HY
15-25% 5 Yr HY
10-15% 5 Yr HY
Index 10 Yr IG
0-10% 5 Yr HY
15-30% 10 Yr IG
10-15% 10 Yr IG
3-7% 10 Yr IG
7-10% 10 Yr IG
Index 5 Yr IG
0-3% 10 Yr IG
15-30% 5 Yr IG
10-15% 5 Yr IG
3-7% 5 Yr IG
7-10% 5 Yr IG
0-3% 5 Yr IG
0.0%
Note: Assuming 40% recovery for IG and 35% recovery for HY; Source: Morgan Stanley
Tranche thickness, expected number of defaults for the index, and presence of a upfront payment are important variables to watch in assessing default sensitivity. A thicker tranche, e.g., high yield equity versus investment grade equity, implies lower sensitivity to a default. Similarly, if the index is expected to have a large number of defaults, a single default has a smaller impact. A tranche with upfront payment has lower default sensitivity, compared to the same tranche without an upfront payment, much like spread sensitivity. Again, the reason for lower sensitivity is that, due to an upfront payment, the protection writer does not lose premium due to a default, while in the absence of an upfront payment, the running premium would decline proportionally to the amount of tranche notional lost due to defaults. Of the three equity tranches we mentioned, the five-year high yield equity tranche is the thickest, has the highest number of expected defaults (given that the high yield index trades much wider), and has all of its premium paid upfront. As a result, it has significantly lower sensitivity to defaults versus the investment grade equity tranches.
Please see additional important disclosures at the end of this report.
75
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 6
How do we determine if a tranche has more default risk or spread risk? Delta-neutral default sensitivity, as shown in Exhibit 13, helps us demarcate the subordination level in the capital structure where tranches transition from being net default sensitive to net spread sensitive. For example, the mezzanine tranche (3-7%) of five-year investment grade is net spread sensitive (i.e., the delta-neutral protection writer will have a positive mark-to-market in case of a default), while the 10-year mezzanine tranche is net default sensitive. exhibit 13
-8%
Delta Neutral P/L Due to 1 Default (% of Notional)
-7% -6% -5% -4% -3% -2% -1% 0% 1% 35-100% 5 Yr HY
25-35% 5 Yr HY
15-25% 5 Yr HY
10-15% 5 Yr HY
0-10% 5 Yr HY
15-30% 10 Yr IG
10-15% 10 Yr IG
3-7% 10 Yr IG
7-10% 10 Yr IG
0-3% 10 Yr IG
15-30% 5 Yr IG
10-15% 5 Yr IG
3-7% 5 Yr IG
7-10% 5 Yr IG
0-3% 5 Yr IG
2%
Note: Assuming 40% recovery for IG and 35% recovery for HY; Source: Morgan Stanley
Utilizing the distinction between net default and spread sensitivities of tranches, we can construct trades that are more efficient in expressing our credit views. For example, a combination of long equity and short mezzanine in a delta neutral ratio effectively expresses a credit view that is constructive on default risk but not on spread levels. “THETA” OR TIME DECAY
As a credit default swap approaches maturity, the spread is bound to converge to zero, since credit protection provided by a CDS eventually becomes worthless if a default does not occur. Furthermore, the rate of decline in the value of protection is determined by the slope of the credit curve. An upward sloping credit curve implies that a larger number of defaults are expected to occur towards the end of the index maturity. For example, the current high yield DJ CDX curve implies that about one-third of the total expected defaults in the coming five years are projected to occur during the fifth year. After one year, assuming no defaults occur during the year and spreads roll down the curve, essentially these defaults disappear, implying a substantial gain for the protection writer. Since junior tranches are levered investments on defaults, their value (to protection seller) rises much faster than the index. Exhibit 14 shows the total return of DJ CDX IG five-year tranches for different time horizons, from the perspective of the protection seller. The value of protection declines precipitously initially, assuming a static credit curve (i.e., spreads ride down the curve with the passage of time), and decelerates toward the maturity. On an absolute basis, tranches with higher than 1x delta lose value
76
faster than index, while senior tranches with lower than 1x delta lose value slower than the index, as one would expect. exhibit 14
Time Decay – 5 Yr IG
Total Return 60%
50%
0-3%
3-7%
7-10%
10-15%
15-30%
Index
40%
30%
20%
10%
0% Current
1 Year
2 Years
3 Years
4 Years
5 Years
Source: Morgan Stanley
However, when we analyze a tranche’s time decay relative to itself over time (i.e., how much of the total value is realized every year), we find that equity tranche value decays slower than the index while all other tranches decay faster than the index, in the case of investment grade five-year tranches. More importantly, about 50% of the total return is realized within the first year for tranches other than the equity tranche (assuming static base correlation and credit curves). For further details, refer to Exhibit 15. exhibit 15
Speed of Time Decay – 5 Yr IG
Proportion of Total Return 100%
75%
50%
25%
0-3%
3-7%
7-10%
10-15%
15-30%
Index
0% Current
1 Year
2 Years
3 Years
4 Years
5 Years
Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
77
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 6
“RHO” OR SENSITIVITY TO CORRELATION CHANGES
As we discussed in Chapter 5, junior tranches are long correlation, while senior tranches are short correlation and mezzanine tranches are relatively insensitive. In other words, higher correlation is better for junior tranches and worse for senior tranches (from the perspective of protection sellers). Since premium is the compensation for taking default risk, the par premium for an equity tranche decreases with rising correlation while it falls for senior tranches. Exhibit 16 shows the sensitivity of five-year investment grade tranches to parallel shifts in the base correlation curve. For example, for the +5% scenario, we assumed that the equity tranche’s correlation increases 5% and the base correlation skew curve remains unchanged. Thus for all tranches, the attachment and detachment point correlations change by the same amount. exhibit 16
Impact of Correlation on Par Spreads – 5 Yr IG
Par Spread (bps) 1,800 1,600
0-3%
3-7%
7-10%
10-15%
15-30%
1,400 1,200 1,000 800 600 400 200 0 -10%
-5%
0% 5% Correlation Curve Shift
10%
15%
Source: Morgan Stanley
Since base correlation considers all tranches as a portfolio of two equity tranches, as the skew (i.e., the difference between the implied correlations for attachment and detachment points) increases, the par spread decreases, just like the equity tranche.
78
exhibit 17
Correlation Sensitivity Changes with Changes in Spreads – 5 Yr IG
Mark-to-Market Impact of 1% Change in Correlation
0-3%
1.0%
3-7% 7-10% 10-15%
0.8%
15-30% 0.6% 0.4% 0.2% 0.0% -0.2% 0
25
50
75 100 125 Index Spread Level (bps)
150
175
200
Source: Morgan Stanley
It is important to note that correlation sensitivity of tranches changes as the underlying index spread moves. In Exhibit 17, we analyze the mark-to-market impact of a 1% increase in correlation for five-year investment grade tranches at different spread levels for the index. As shown in the exhibit, senior tranches are relatively correlation insensitive for small changes in spread but become significantly more sensitive as the index widens materially. The converse is true for the equity tranche, i.e., it is rather sensitive for small changes to the index, but beyond a certain point of widening, correlation sensitivity starts to decline. To isolate the impact of correlation changes, we have assumed that the base correlation skew remains unchanged. However, given a large number of ratings-based investors in very senior tranches, these tranches have a non-zero floor on price. In other words, when spreads fall to very low levels, the pricing on very senior tranches is more resilient. Consequently, base correlation skew tends to rise as spreads fall to very low levels and to flatten as spreads increase. CONCLUSION
In this chapter, we discussed tranche greeks from a conceptual perspective. We find that thinking of tranches as options on defaults helps one develop intuition for the various sensitivities. Spread sensitivity (delta) declines as we go higher up in the capital structure; however, upfront payments tend to decrease this sensitivity. Furthermore, tranche deltas themselves are not constant, and change with changes in the index and the elapse of time. In-the-money tranches have favorable convexity for protection sellers, while convexity in out-of-the-money tranches favors protection buyers. Time decay for tranches is very rapid in the first few years and decelerates materially as they approach maturity. Finally, subordinate tranches benefit from rising correlation, while senior tranches lose value and mezzanine tranches are relatively insensitive.
Please see additional important disclosures at the end of this report.
79
Section B
Valuation and Sensitivity
Please see additional important disclosures at the end of this report.
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 7
A Framework for Secondary Market CDO Valuation
October 1, 2001
Primary Analyst: Sivan Mahadevan David Schwartz
Introduction As we watch the evolution of a secondary market in CDO notes and equity, we focus our research efforts on understanding how CDO valuations can be influenced by the multitude of factors that affect the value of these securities and their underlying collateral. A collateralized debt obligation is effectively an investment in a leveraged portfolio of credit instruments. The risk of this portfolio is disproportionately redistributed into the various tranches issued by the special purpose vehicle administering the investment portfolio. The principal risks investors face are price volatility, liquidity and default risk, and there are many metrics that measure these characteristics. Investors purchase CDO notes for a variety of reasons. The senior notes of a CDO offer good yield pickup over similarly rated instruments in the cash credit markets. These notes earn their high ratings because the CDO cash flow structure offers them substantial protection from defaults in the underlying collateral. The subordinated notes and residual components (equity) of CDOs offer investors unique risk and return characteristics not easily replicated in the cash or credit derivatives markets. Most CDOs are structured so that if default and recovery rates over their lifespan fall within a range that includes average historical rates, investors stand to benefit from the leverage of the structure and the resulting higher yields the CDO tranches offer relative to investments in the corporate bond markets. Many traditional CDO note and equity holders have been buy and hold investors. The bulk of arbitrage CDOs are structured as cash flow CDOs, meaning that the collateral managers focus their efforts on meeting the cash flow liabilities while conforming to the structural guidelines of the various tranches. In cash flow deals, collateral managers’ concerns about the market value or the price volatility of the underlying collateral is generally in the context of how such valuations can affect the credit quality of the collateral or the cost of transactions. This style is similar to asset/liability investment managers such as insurance companies. In arbitrage market value CDOs, the collateral managers focus on improving the market value of the underlying collateral, similar to how total-return oriented investment managers operate. Independent of the structure and collateral management style of CDOs, buyers and sellers of CDO notes and equity in the secondary market are interested in determining a fair (and agreed upon) value for the expected cash flows, price volatility, liquidity, and default risk that a given CDO tranche represents. The goal of this chapter is to discuss objective ways for valuing CDO tranches in the secondary market, and to apply subjective considerations including market technical information.
82
In this chapter, we present several methodologies for secondary market CDO note and equity valuation with a focus on arbitrage cash flow CDO structures. The approaches differ in degrees of computational complexity and required market savvy, but ultimately they all help investors gauge the value of secondary market CDO investment opportunities.
CDO Valuation Framework Before we describe a valuation framework for CDOs, we make some observations on how related financial instruments are valued in the secondary markets. Investment grade corporate bonds are traditionally valued on a comparables basis. Liquid bonds from an issuer are compared to liquid bonds from other issuers with similar characteristics. The characteristics considered include credit quality, industrial sector, and maturity, and from these comparisons, spread levels are determined. 1 Bonds that have unique characteristics (such as optionality or special credit considerations) move away from their comparables, and spreads are adjusted accordingly, either independently, with respect to a new set of comparables, or with some sophisticated valuation techniques. For example, bonds with options can be valued in the HeathJarrow-Morton framework using a multi-path simulation. Higher rated non-investment grade corporate bonds can be valued on a comparables basis as well, but reliance on these types of comparisons quickly disappears in favor of unique valuations derived from credit fundamentals and investor appetite. Valuation techniques for asset backed securities vary given potential differences in security structure, but the comparables technique for securities with similar collateral and structure is common. Valuation techniques for collateralized mortgage obligations (CMOs) are worth observing, given their conceptual similarity to CDOs. The least risky (i.e., most immune to prepayment risk) tranches of CMOs tend to be tested using various metrics and then compared to other similar lower-risk CMO tranches. The residual components of CMOs (known as Z-bonds, which carry the bulk of the collateral’s prepayment risk) are often valued using modeling techniques that generate and discount cash flows over multiple input paths and average these results. With this background in mind, we begin determining a framework for valuing CDO notes and equity by asking three fundamental questions: •
What common market practices exist today for valuing CDOs?
•
What relevant input from the financial markets should influence CDO valuations?
•
What financial valuation techniques can be applied to CDOs?
The answers to the above questions led us to organize our CDO valuation framework along three basic methodologies. 1
Tracking and Outperforming Bond Indices: Morgan Stanley Fixed Income Research, July 9, 2000, R. Fuhrman, et al.
Please see additional important disclosures at the end of this report.
83
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 7
THREE VALUATION METHODOLOGIES
Market practices for determining secondary market CDO note and equity prices exhibit a range of required market savvy and computational complexity. The computationally simpler approaches involve generalizing collateral characteristics and comparing CDO tranches to similar structures with a common credit rating and collateral type. Computationally more complex approaches exist as well, ranging from valuing tranches over a handful of input paths to full-blown simulation models run over a large distribution of paths. In this chapter, we describe three methodologies for valuing CDO notes and equity in the secondary market. •
Re-rating Methodology, an approach that is computationally relatively simple and involves inferring a rating for a CDO tranche through a single set of assumptions and comparing it to other similarly rated tranches to derive a value.
•
Market Value Methodology, a technique that involves equating the market value of the assets of a CDO (the collateral) with the market value of the liabilities (debt, equity and management fees). Market value changes can be indicators of potential structural changes in a CDO as well.
•
Cash Flow Methodology, a technique that is based on financial engineering fundamentals, namely generating and discounting tranche cash flows along input paths. Variants of this approach include a single assumption set, several assumption paths, and a simulation approach over a large distribution of events.
Exhibit 1 summarizes the spectrum of inputs and valuation methodologies for determining CDO tranche values.
84
Please see additional important disclosures at the end of this report.
85
exhibit 1
Source: Morgan Stanley
Secondary Market CDO Valuation Methodologies
Re-rating Methodology Adjust risk characteristics, calculate expected loss, compare to similar structures
CDO Deal and Collateral Info
or
Process Inputs
Determine Value
Technical Factors
Market Value Methodology Use asset and liability equivalence relationship as an indicator of value
or
Pricing Methodologies
Inputs
Cash Flow Methodology Generate and discount cash flows over one or multiple paths, or through a simulation
Default Rates and Default Correlation
CDO Primary and Secondary Markets
Make Assumptions
IG/HY Credit Markets
Default Swap Markets
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 7
The Re-rating Methodology The re-rating methodology is based on a simple, quantitative approach to identifying a credit rating for a CDO note. Once a credit rating is identified, the note is compared to similar structures in the marketplace to determine a spread value. Comparability is usually based on CDO characteristics such as collateral type and cohort year of issuance. The approach is particularly valuable when an investor has newly updated information on a CDO, or would like to forecast changes to a CDO’s structure. In such cases, the credit rating calculated by the re-rating methodology can be different from those assigned by ratings agencies, giving the investor insight into the current structure or evolution of the CDO. The approach is also useful in providing investors with a guide to how sensitive a tranche’s credit rating is to changes in underlying assumptions. The re-rating methodology is based on the well-known Moody’s Binomial Expansion Method for determining a CDO tranche rating. 2 The Moody’s approach is a simple method that involves generalizing a CDO’s collateral pool, inferring a default rate for this portfolio and calculating expected losses for a given CDO tranche, which leads to a credit rating using Moody’s association of default rates and ratings. The re-rating methodology is described in the flow chart in Exhibit 3. The expected loss of a CDO tranche is computed based on two inputs: the binomial distribution of defaults and the loss distribution of a CDO tranche. From these two functions, we calculate the probability-weighted loss (the expected loss) for the tranche, which can then be mapped to a Moody’s rating using the Moody’s “Idealized” Cumulative Expected Loss Rates.2 GENERALIZING THE COLLATERAL – THE DIVERSITY SCORE
The re-rating methodology gains its simplicity from Moody’s approach to generalizing the underlying CDO collateral. Moody’s simplifies the collateral portfolio by breaking it down into a set of uncorrelated assets. Correlation is defined as default correlation (not correlation of asset returns or changes in spreads). Given a portfolio of correlated assets, the diversity score represents the equivalent number of uncorrelated assets. If a portfolio has 30 uncorrelated assets, then it has a diversity score of 30. How does Moody’s simplify a portfolio to a collection of uncorrelated assets? The basic assumption is that assets in different industry categories are uncorrelated. If a portfolio has 30 credits (each from a different Moody’s industry category) then the portfolio has a diversity score of 30. Each of the 30 represented industry categories contributes 1.0 units to the total diversity. If more than one credit from the portfolio belongs to a given industry category, then that category’s contribution to the total increases, following a sliding scale depicted in Exhibit 4. This relationship implies that credits in the same industry group have a positive default correlation. In Exhibit 5, we show that the implied default correlation between credits in the same industry falls from 33.3% to 16.7% as the number of credits in the industry increases from 2 to 10.
2
The Binomial Expansion Method Applied to CBO/CLO Analysis, Moody’s Investors Services, Dec 13, 1996.
86
exhibit 2
Moody’s Industry Categories
Number 1
Industry Aerospace & Defense
2
Automobile
3
Banking
4
Beverage, Food & Tobacco
5
Buildings and Real Estate
6
Chemicals, Plastics & Rubber
7
Containers, Packaging and Glass
8
Manufacturing
9
Diversified/Conglomerate Manufacturing
10
Diversified/Conglomerate Service
11
Metals & Minerals
12
Ecological
13
Electronics
14
Finance
15
Farming and Agriculture
16
Grocery
17
Healthcare, Education and Childcare
18
Home and Office Furnishings, Housewares and Durable Consumer Products
19
Hotels, Inns and Gaming
20
Insurance
21
Leisure, Amusement, Motion Picture, Entertainment
22
Machinery
23
Mining, Steel, Iron and Nonprecious Metals
24
Oil and Gas
25
Personal, Food and Miscellaneous Services
26
Printing and Publishing
27
Cargo Transport
28
Retail Stores
29
Telecommunications
30
Textiles and Leather
31
Personal Transportation
32
Utilities
33
Broadcasting
Source: Moody’s
Please see additional important disclosures at the end of this report.
87
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 7
exhibit 3
WARF
The Re-rating Methodology
Recovery Rates
Diversity
Default Timing
Collateral Par Changes
Use Cash Flow Waterfall to Generate Loss Distribution
Use Binomial Method to Generate Default Probability Distribution
Calculate Expected Loss and Implied Default Probability Map Default Probability to Credit Rating
Determine Value
Source: Morgan Stanley
exhibit 4
4.0 3.5
Diversity Score
Moody’s Diversity Score Contribution for One Industry
3.0 2.5 2.0 1.5 1.0 0.5 0.0 1
2
3
4
5
6
7
8
9
10
Number of Assets in Same Industry
Source: Moody’s
exhibit 5
0.35 Implied Default Correlation
Implied Default Correlation
0.30 0.25 0.20 0.15 0.10 0.05 0.00 2
3
4
5
6
7
8
9
10
Number of Assets in Same Industry
Source: Morgan Stanley, Moody’s
The implied correlation calculation is based on moment-matching, with the assumptions that each credit has the same weight in the portfolio and that the default correlation between any two assets in the same industry are equal.
88
MODELING DEFAULTS AS A BINOMIAL DISTRIBUTION
Moody’s uses the diversity score of a CDO to model the default behavior of the collateral. The approach assumes that the number of defaults in a portfolio of N uncorrelated assets (i.e., diversity score = N) follows a binomial distribution. The binomial distribution is used as follows. First, an event of default is considered to be a binary operation: an asset either experiences default, or it does not. The probability of an asset defaulting is represented by p . The probability of an asset not defaulting is equal to 1− p . The Moody’s method assumes that all assets in a portfolio have the same probability of default, derived from the WARF (numerical weighted-average ratings factor for the collateral). Given these simplifications, we can model the default behavior of a portfolio using a binomial tree. Consider a portfolio of three assets. There are several possible default behaviors for this portfolio: the portfolio can experience zero defaults, one default, two defaults, or three defaults. The binomial tree in Exhibit 6 depicts all of the possible default outcomes. exhibit 6
3
p
Binomial Tree of Defaults for a Three Asset Portfolio
p 1-p 2
p
p 1-p p
1-p
1-p
1 p 1-p 1-p
0
Source: Morgan Stanley
There are eight possible paths in this binomial tree, and they are summarized in Exhibit 7. exhibit 7
Number of Defaults 0 1
Distribution of Defaults for a Three Asset Portfolio Frequency of Occurrence in Binomial Tree 1 3
Probability 3 (1-p) p(1-p)
2
2
2
3
P (1-p)
3
1
p
3
Weighted Prob 3 (1-p) 3p(1-p)
2
2
3p (1-p) 3
p
Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
89
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 7
Given this binomial distribution-based representation of portfolio default behavior, we can compute the portfolio’s default probability distribution. The probability that a portfolio of N (uncorrelated) assets experiences j defaults is:
§N· Pj = ¨¨ ¸¸ p j (1 − p ) ( N − j ) © j¹ The first parenthetical part of this expression is a combination, or the number of groupings of j defaulted assets from a portfolio of N total assets (j <= N). This is mathematically equal to
§N· N! ¨¨ ¸¸ = © j ¹ j!( N − j )! The notation N ! (read “N factorial”) is equivalent to Nx( N − 1) x( N − 2) x...x1 .The variable p is the generalized cumulative probability of default for any asset in the portfolio (derived from the portfolio WARF). exhibit 8
p = 0.5
p = 0.15
25% Probability
Binomial Distribution of Defaults for a 20 Asset Portfolio
30%
20% 15% 10% 5% 0% 0
5
10
15
20
Number of Defaults
Source: Morgan Stanley
The graph in Exhibit 8 shows the binomial distribution of defaults for a hypothetical 20-asset portfolio. The two curves represent different default probabilities, p=0.5 and p=0.15. Does it make sense that the default behavior of a portfolio follows a bell-shaped curve? What the curve tells us is that for very risky credits with a cumulative default probability of 50% over the life of the structure, the probability of having zero defaults (out of 20) is close to 0%. The probability of having one default out of 20 is higher (but still very small), and the probability rises until you reach 10 defaults, after which it falls to near zero values as you approach a fully defaulted portfolio. In other words, the probability of experiencing only very few defaults is small, and the probability of experiencing a large number of defaults is also small – but the probability of having 10 defaults is the highest (17.6%) over the life of the structure. In general, the binomial distribution is not a symmetric distribution, and the degree to which it is skewed is related to the probability of default of the individual assets and
90
the number of assets. The distribution is symmetric if p = 0.5. For any p, the distribution becomes more symmetric as the number of assets increases. COMPUTING A CDO TRANCHE’S EXPECTED LOSS
We can use the CDO’s cash flow waterfall to determine the loss a CDO tranche will experience for a given number of defaults (see Chapter 1). In the example in Exhibit 9, the diversity score is 35 and p = 21.5%. The bell-shaped graph and the left-hand scale show the binomial distribution. The columns and right-hand scale show the losses that the hypothetical CDO tranche will experience under different default scenarios. We can now combine this tranche loss distribution with the binomial distribution to calculate the probability-weighted loss, otherwise known as the expected loss. N
Expected Loss = ¦ Pj L j j =0
The variable Pj is the probability that the collateral portfolio experiences j defaults and the variable Lj represents tranche’s loss when the collateral experiences j defaults. exhibit 9
100% 90%
70% 60% 10%
50%
Tranche Loss
80% 15%
Probability
Binomial Distribution and Tranche Loss Distribution
20%
40% 30% 5% 20% 10% 0%
0% 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35
Number of Defaults
Source: Morgan Stanley
INFERRING A MOODY’S CREDIT RATING
We can use the CDO tranche’s expected loss to infer a Moody’s credit rating. To do this, we must first assume a recovery rate for defaulted securities. Given this recovery rate, we can compute the implied default probability of the CDO tranche as follows: Implied Def Prob = Expected Loss/(1-Recovery Rate) We can map the implied default probability to a Moody’s credit rating using the Moody’s “Idealized” Cumulative Expected Loss Rates (see Exhibit 10). Note that this mapping is usually not done directly. Instead, implied default probabilities are multiplied by a stress factor (dependent on the target credit rating) before mapping a credit rating, adding some cushion to the inferred credit ratings. Finally, given this rating, a value for the CDO tranche can be determined from a comparables analysis as discussed at the beginning of this section.
Please see additional important disclosures at the end of this report.
91
92 11.670 16.610
2.810 4.680 7.160 11.620 26.000
B1
B2
B3
Caa
Source: Moody’s
3.470
1.560
Ba2
32.500
8.380
5.510
1.050
0.470
0.280
Ba3
0.090
Baa1
0.150
2.020
0.039
A3
0.070
0.037
0.870
0.011
A2
Ba1
0.006
A1
0.019
0.170
0.003
Aa3
0.003 0.008
0.420
0.001
Aa2
Baa3
0.001
Aa1
2 Years 0.000
Baa2
1 Year 0.000
39.000
21.030
15.550
11.580
7.870
5.180
3.130
1.710
0.830
0.560
0.360
0.222
0.117
0.059
0.026
0.010
3 Years 0.001
43.880
24.040
18.130
13.850
9.790
6.800
4.200
2.380
1.200
0.830
0.540
0.345
0.189
0.101
0.047
0.021
4 Years 0.002
48.750
27.050
20.710
16.120
11.860
8.410
5.280
3.050
1.580
1.100
0.730
0.467
0.261
0.142
0.068
0.031
5 Years 0.003
52.000
29.200
22.650
17.890
13.490
9.770
6.250
3.700
1.970
1.370
0.910
0.583
0.330
0.183
0.089
0.042
6 Years 0.004
55.250
31.000
24.010
19.130
14.620
10.700
7.060
4.330
2.410
1.670
1.110
0.710
0.406
0.227
0.111
0.054
7 Years 0.005
58.500
32.580
25.150
20.230
15.710
11.660
7.890
4.970
2.850
1.970
1.300
0.829
0.480
0.272
0.135
0.067
8 Years 0.007
61.750
33.780
26.220
21.240
16.710
12.650
8.690
5.570
3.240
2.270
1.520
0.982
0.573
0.327
0.164
0.082
9 Years 0.008
65.000
34.900
27.200
22.200
17.660
13.500
9.400
6.100
3.600
2.600
1.800
1.200
0.700
0.400
0.200
0.100
10 Years 0.010
Moody’s Idealized Cumulative Probability of Default by Rating and Number of Years
chapter 7
Rating Aaa
exhibit 10
Structured Credit Insights – Instruments, Valuation and Strategies
TRANCHE RATING SENSITIVITY
Through implied default probability mappings, the graphs in Exhibits 11-13 show the rating sensitivity of a hypothetical CDO tranche to changes in portfolio diversity score, WARF, and par loss (realized defaults). Such sensitivity analysis is an important use of the re-rating methodology as it gives investors insights into the stability of current or forecasted ratings. For example, a change in the collateral WARF from 1850 to 1950 will move the hypothetical tranche from an investment grade Baa3 rating one notch lower to Ba1. Similarly, an immediate par loss increase from 6% to 7% will lower the rating one notch from Ba1 to Ba2. exhibit 11
Tranche Ratings Sensitivity to Collateral Diversity Score
Implied Default Probability 14% Ba3 Ba2 Ba1 Baa3
12% 10%
Baa2 Baa1 A3
8% 6% 4% 2% 0% 15
20
25
30
35 40 Diversity Score
45
50
55
60
2,050
2,150
Source: Morgan Stanley
exhibit 12
Tranche Ratings Sensitivity to Collateral WARF
Implied Default Profitability 14% Baa2 Baa1 A3
Ba3 Ba2 Ba1 Baa3
12% 10% 8% 6% 4% 2% 0% 1,350
1,450
1,550
1,650
1,750 WARF
1,850
1,950
Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
93
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 7
exhibit 13
Tranche Ratings Sensitivity to Immediate Collateral Par Losses
Implied Default Probability
Ba3 Ba2 Ba1 Baa3
16%
Baa2 Baa1 A3
14% 12% 10% 8% 6% 4% 2% 0% 0
1
2
3
4 5 Par Loss (%)
6
7
8
Source: Morgan Stanley
VALUING EQUITY AS THE RESIDUAL
The re-rating methodology, by definition, is useful only for determining value of CDO liabilities that carry a credit rating. How does one use this methodology to determine value for equity tranches that are not rated? The answer is that we can combine the valuations of rated CDO tranches in this approach with other methodologies (described in the following sections) to determine the residual value of the equity tranche.
The Market Value Methodology One simple view of a CDO is to think of it as a portfolio with cash flows divided into many pieces. The sum of these parts must equal the whole portfolio. This equivalence property is the basis of the CDO valuation technique we call the market value methodology, the fundamental premise of which is: Market Value of Assets = Market Value of Liabilities The assets of a CDO are its collateral (adjusted for any hedges that may be in the structure). The liabilities of a CDO are the notes issued by the special purpose vehicle, the equity and any fees collected by collateral managers (over the life of a CDO) and by underwriters and administrators (at issuance). When CDOs are issued, this equivalence relationship between assets and liabilities exists (usually within a small band), but as CDOs age, we have found that the market value of assets does not always stay in line with the market value of liabilities. Absent are strong forces to drive equivalence, including: •
The difficulty of going long one side of the asset liability equation and short the other. Arbitrageurs are not easily able to drive the equivalence relationship.
•
The buy and hold behavior of traditional CDO note and equity investors.
It should be noted that closed-end mutual funds experience similar behavior in the relationship between the NAV of the fund and the value of the fund’s shares. Nevertheless, market value equivalence is a fundamental relationship that can be used as a relevant metric in valuing CDO liabilities. Furthermore, we find that many CDO
94
market participants rely on the market value equivalence, and in this section we describe a market value based approach for valuing CDOs. THE ROLE OF CHANGING MARKET VALUE
In theory, whenever we observe a change in the market value of the assets we should observe an equivalent change in the value of CDO liabilities. For example, a 10 bp widening in credit spreads of a CDO’s underlying collateral should have a commensurate impact on the market value of the CDO’s notes and equity. How can we measure this impact? The market value approach tells us that a decrease in value due to a 10 bp widening in the credit spread of the underlying collateral should be distributed to the various CDO tranches such that: •
Each tranche is affected in a fair way.
•
The market value of assets continues to equal the market value of liabilities.
How do we determine a fair way of distributing the change in market value to the various tranches? One extreme is for the tranches to absorb the losses sequentially. The most subordinated tranche (equity) would absorb losses until it lost its full par value, at which point the next subordinated tranche would begin to absorb losses. Another extreme is to distribute the loss in a pro-rata format to all tranches simultaneously (see Exhibit 14).
Please see additional important disclosures at the end of this report.
95
96
0
5
15
Seniors
Seniors Mezzanine Equity
= 100% = 100% = 100%
20
= 100% = 100% = 0%
Collateral Loss (%)
10
Mezzanine
Equity
Seniors Mezzanine Equity
25
Seniors Mezzanine Equity
Sequential Approach
30
= 93% = 0% = 0%
0
5
Seniors Mezzanine Equity
Mezzanine
Equity
= 100% = 100% = 100%
10 15 20 Collateral Loss (%)
Seniors
Seniors Mezzanine Equity
Pro-Rata Approach
25
30
Seniors Mezzanine Equity
= 90% = 90% = 90% = 70% = 70% = 70%
chapter 7
Source: Morgan Stanley
Market Value Impact: Sequential and Pro-Rata Approaches
exhibit 14
Structured Credit Insights – Instruments, Valuation and Strategies
Market Value
Market Value
Consider a simple hypothetical example, a CDO with three tranches as depicted in Exhibit 15. If the collateral loses 10% of its value, under the sequential approach, the equity tranche will absorb the entire loss (its price would fall to 0% of par, as equity comprises 10% of the par value of the structure). The prices of the other more senior tranches would remain at 100% of par. Under the pro-rata approach, the loss would be distributed equally to each tranche (taking into consideration the tranche’s weight within the CDO). The price of each tranche would therefore fall to 90% of par. With a 30% loss of collateral market value, under the sequential approach, both subordinate tranches would lose all of their value (given that they make up 25% of the structure combined) and the remaining loss would be applied to the senior tranche, bringing its price as a percentage of par to 93.3. Under the pro-rata approach, each tranche would lose 30% of its value, so all tranches would be priced at 70% of par. exhibit 15
Tranche Senior Mezzanine Equity
Original Price 100 100 100
Sequential and Pro-Rata Approaches: Examples
% of Par Value 75 15 10
10% Collateral Loss Pro-Rata Sequential Approach Approach (Price) (Price) 100 90 100 90 0 90
30% Collateral Loss Sequential Pro-Rata Approach Approach (Price) (Price) 93 70 0 70 0 70
Source: Morgan Stanley
Is either extreme approach correct? Neither accurately reflects the risk of the various tranches, but the real values are likely bounded by these two approaches. For moderate declines in collateral value, the sequential approach where the subordinate tranche absorbs the entire loss is probably closer to what is practiced in the marketplace. In our view, to understand how the various tranches are affected, it is important to consider both the bounds and the nature of the market value changes. A PRACTICAL APPROACH: CATEGORIZING MARKET VALUE CHANGES
As we mentioned in our introduction, collateral managers of cash flow CDOs are primarily concerned about meeting the cash flow liabilities of the CDO, while keeping the CDO in compliance with credit ratings, par coverage and interest coverage guidelines. Changes in market value of the underlying collateral are not necessarily of concern to collateral managers, unless these changes are large enough or specific enough in nature to have a structural impact on a cash flow CDO. Examples of changes in collateral that structurally affect a CDO include credit rating upgrades or downgrades, defaults, redemptions and reinvestment. From the perspective of cash flow CDOs, we can think of collateral market value changes as belonging to one of three categories: •
Small market value changes that will not lead to structural changes in cash flow CDOs. This includes small changes in the general level of interest rates and credit spreads.
•
Large market value changes that may not lead to structural changes in cash flow CDOs, but are large enough to change assumptions about the general level of interest rates and credit spreads.
•
Market value changes that are indicators of potential structural change in a CDO.
Please see additional important disclosures at the end of this report.
97
98 Credit specific moves that are potential indicators of changes in ratings, defaults, or triggers that will force transactions or cash flow diversions
Market value changes that indicate structural changes in the CDO
Source: Morgan Stanley
Large moves in the general level of interest rates or credit spreads
Description Examples include changes in general level of interest rates or credit spreads
Value tranche using rating or cash flow methodologies, taking collateral indicators into consideration
Market savvy required, but some of the impact should be absorbed here. Observe new issue comparables
Likely Impact on Senior Notes None
Value tranche using re-rating or cash flow methodologies, taking collateral indicators into consideration
Market savvy required, but the impact should be significant. Observe new issue comparables
Likely Impact on Second Priority or Mezzanine Notes Market savvy required. Observe new issue comparables
Value tranche using re-rating (residual) or cash flow methodologies, taking indicators into consideration
Change discount rate assumptions (see cash flow methodology)
Likely Impact on Equity Market savvy required, mark up or down on a price basis
A Practical Approach to Understanding the Impact of Collateral Market Value Changes
chapter 7
Large market value change with no structural CDO impact
Nature of Market Value Change Small market value change with no structural CDO impact
exhibit 16
Structured Credit Insights – Instruments, Valuation and Strategies
In Exhibit 16, we describe these three market value classifications along with the likely impact based on what we have observed in the market place. It is important to note, however, that the extent to which senior or mezzanine tranches are affected will depend on the amount of losses already incurred by the more subordinate tranches. For small changes in market value that are not indicators of future structural changes in the CDO, we have observed that the senior notes are likely unaffected, while the equity and mezzanine notes typically absorb the bulk of the impact (if any). A guide to the magnitude of this impact can be the new issue CDO market, where investors observe the value of comparable CDO tranches. If there are large changes in market value that are not indicators of future structural changes in the CDO, then we have observed all tranches of a CDO being impacted. Senior notes generally take their cue from comparables or the general level of high credit quality instruments in the cash markets. Market savvy is required for the second priority or mezzanine notes, just as we noted before. For equity, large changes in interest rates and credit spreads should influence fundamental assumptions like the yield (discount rate) required by investors (see the cash flow methodology). MARKET VALUE AS AN INDICATOR OF STRUCTURAL CHANGE
The most insightful use of the market value approach is to use changes in market value as an indicator of potential structural change in a CDO. Examples of these indicators include: •
The market prices in a ratings upgrade or downgrade for a particular credit.
•
The market expects a credit to default or perceives an increased probability of default.
•
Collateral manager may be forced to sell a credit that falls below a ratings threshold (e.g., CCC).
•
Collateral manager may change a view on a particular credit based on information that the market has already priced in which can trigger a sale.
Each of these examples gives the investor an indication of the potentially changing characteristics of a CDO as a whole or a specific tranche. Investors can then use other methodologies (i.e., the re-ratings or cash flow methodologies) to value the tranche. For example, if the market begins to price in an upgrade or downgrade for a particular credit, an investor could forecast a different WARF for the collateral, and use the rerating methodology to value the tranche in question. If the market expects a credit to default (or if a credit falls to a ratings threshold that may force the collateral manager to sell it), then the investor can forecast a collateral sale and reinvestment and then revalue the tranche using the cash flow methodology.
The Cash Flow Methodology The cash flow approach draws its support from one of the most fundamental valuation techniques for fixed income securities: generating cash flows and computing their present value using yield or a series of discount factors. Rated notes of CDOs have scheduled cash flows, so in theory, valuing rated notes by generating and discounting cash flows seems reasonable. CDO equity does not have contractually obligated cash
Please see additional important disclosures at the end of this report.
99
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 7
flows, but for a given set of assumptions, there are definitive cash flows that can be valued as well. VALUING CASH FLOWS OF RATED CDO NOTES
What factors influence the cash flows of a rated note? For cash flow CDOs, default and recovery assumptions for the underlying collateral are clearly important. The underlying collateral for most CDOs is not static, so investment and reinvestment strategies and guidelines are important as well. Furthermore, any reasonable cash flow valuation based on default, recovery and reinvestment assumptions would require some measure of sensitivity to changes in these assumptions (that is, investors rarely have one view on these assumptions, and would rather see how altering assumptions can affect the value). In this section we describe two approaches to valuing a CDO tranche using cash flows. The first approach involves identifying a discount rate (yield) from comparable securities and testing its sensitivity to changes in input assumptions. The second approach involves projecting default-adjusted cash flows and discounting them at Libor over both a single path of assumptions and a distribution of assumptions. THE CASH FLOW APPROACH USING A COMPARABLE YIELD
From a corporate bond perspective, the discount rate we are most familiar with is yield. As we discussed earlier, corporate bond investors often value credits on a comparables basis, or making an educated guess at the right yield (based on comparable credits). In valuing CDO tranche cash flows, this comparables basis is used as well, along with tests of sensitivity to changes in input assumptions A CDO note’s spread is most likely based on spread levels for comparable notes (new issue or secondary market) and possibly includes other market knowledge about the specific structure. However, the heart of the approach lies in testing the sensitivity of the note’s present value of cash flows (given a yield) to changes in underlying assumptions. Consider the example in Exhibit 17. If we assume that second priority notes on investment grade collateral CDOs of recent vintage have a yield of 8%, then we can hold this yield constant and test the sensitivity of the present value of cash flows of the tranche to a set of default and recovery rate assumptions. exhibit 17
PV
Mezzanine Note PV of Cash Flows Sensitivity
120 110 100 90 80 70 60 50 40 30 20 10 0
Recovery Rate = 40
0%
1%
2%
3%
4%
5%
Recovery Rate = 20
6%
7%
Default Rate
Source: Morgan Stanley
100
8%
9% 10% 11% 12%
Please see additional important disclosures at the end of this report.
101
exhibit 18
Source: Morgan Stanley
Cash Flow Methodology Default Rates
Reinvestment Rates Collateral Changes
Discount Cash Flows
Discount Rates
Determine Value
Generate Cash Flows
Assemble Multiple Input Paths
Recovery Rate
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 7
For a recovery rate assumption of 40%, the PV is very stable for annual collateral default rates of 0% to 8%. At a recovery rate of 20%, the stability holds until approximately 6% annual default rate. If these default and recovery rates are within ranges that cover collateral credit quality (with some degree of stressing), then we can say that the tranche’s PV is stable at a yield level of 8%. HOW DO WE DISCOUNT EQUITY CASH FLOWS?
As we described earlier, the equity tranche of most CDOs do not have contractually obligated cash flows, and most don’t carry credit ratings either. However, for a given set of assumptions we can generate definitive cash flows for equity, so we can follow a similar cash flow discounting method as we do for rated notes. We can consider deriving a yield for equity from comparable securities. However, it is hard to find comparables for equity tranches, as many secondary market equity tranches are unique. But referencing comparables is still an option, particularly for more recently issued deals (that could have similar collateral composition). The more common approach, however, continues to be driven by what the market requires for the yield of CDO equity. Many deals are structured to have internal rates of return in the 15-20% range if default and recovery assumptions follow historical averages for a given collateral type. As such, after computing equity cash flows under typical collateral default rate assumptions, discount rates of 15-20% are commonly used to compute present values for equity cash flows. PROJECTING DEFAULT-ADJUSTED CASH FLOWS
The first approach to using cash flows of a CDO note gives us a price only after we determine a value for the yield. How can we determine the price without assuming a yield? A simplified approach to valuing a security is to assume that an investor receives the expected cash flows of a security (adjusted for the default probability – see Exhibit 19). exhibit 19
Remaining Cash Flow
Default-Adjusted Cash Flows Cash Flow
Default Adjustment
Time
Source: Morgan Stanley
We can calculate the expected cash flows by adjusting the scheduled cash flows with the probability of default (see Exhibit 19). We then discount each adjusted cash flow by zero-coupon Libor to compute a present value.
102
PV =
Σ
DefaultProbCashFlow (1+Li)i
We use zero-coupon Libor as the discount factor. In theory, the resulting present value is the price an investor would be willing to pay for a security with cash flows that have been adjusted for expected losses from defaults. We can now use this resulting present value to solve for the appropriate spread (or yield) on the original security (with the original scheduled cash flows).
PV =
Σ (1+(TsyYld + Spread)) OrigCashFlow
i
The above technique gives the default-probability-implied spread, based on an assumed probability of default. The probability of default can be inferred from the default swap market as well. When compared to historical default probabilities, the observed level in the default swap market probably more accurately reflects investors’ views on a credit. What do we do if there is there is no liquid default swap market to reference? To answer this question, we have to consider arbitrage forces. In corporate bond markets, investors can create a default-risk free credit instrument by financing a corporate bond (via an asset swap structure) and then buying default protection in the default swap market. In theory, such an investment structure can drive the relationship between the asset swap spread and default swap spread to near equivalence. In practice, there is a basis between the two values to explain differences due to supply and demand, counter party credit quality and other technical factors. Default Swap Spread = Asset Swap Spread + Default Swap Basis. Is the simple one-path approach to valuing a CDO note described above valid? The approach of default-adjusting cash flows and discounting by zero-coupon Libor is used, but generally in a multi-path context which we describe in the next section. CASH FLOW SIMULATION METHODOLOGY
We have discussed above computing a default-probability-implied yield for a tranche, given one path of input assumptions. We can extend this to a formal simulation where a large distribution of assumptions are generated programmatically, where for each path, cash flows are generated, adjusted based on default and recovery rate assumptions and discounted using risk free rates. The average of these results can be considered a fair value for a CDO note. CASH FLOW SIMULATION APPROACH
For each of thousands (or millions) of possible default scenarios, we repeat the following procedure: •
For each period, determine which individual bonds in the collateral pool default. Recovery and reinvestment assumptions need to be applied to defaulted bonds.
Please see additional important disclosures at the end of this report.
103
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 7
•
Use the cash flow waterfall to determine what payments are made to each tranche in each period.
•
Discount these cash flows at zero-coupon Libor.
Having repeated this process over many different default possibilities, we can average the results for each tranche to determine a fair value. IMPLEMENTING THE SIMULATION: MODELING THE TIMING OF DEFAULTS
In theory this simulation is simple, but the implementation can be quite challenging. First, it can be computationally time consuming to simulate whether or not an individual bond defaults in each short interval of time. To overcome this computational burden, practitioners tend to directly model the amount of time it takes until a given bond defaults. The timing of defaults, not just the number of defaults, is critical for valuing CDOs. Consider the defaults that are graphically depicted in Exhibit 20. exhibit 20
Default Clustering Examples
Default Example 3
2
1
0 Time
Source: Morgan Stanley
There are many ways a portfolio can experience 10 defaults. One extreme is that they are clustered early in the life of a CDO, while another extreme is that they are clustered late in the CDO life. The timing of these defaults affects the cash flow stream available to the various CDO tranches. Many early defaults can trigger the delevering of the structure by senior note holders, resulting in reduced cash flows (or no cash flows at all) to the subordinated note or equity holders. IMPLEMENTING THE SIMULATION: MODELING DEFAULT PROBABILITY AND DEFAULT CORRELATION
To correctly value a CDO, we need to use market-implied (or “risk neutral”) default probabilities. A term structure of default probabilities can be implied from default swap spreads or asset swap spreads. Practitioners tend to model default probabilities using a stochastic process, which can be calibrated to default swap spreads or asset swap spreads.
104
The correlation of defaults must be accounted for as well. Correlation is critical, since individual bonds are affected by a number of common factors (see discussion in the Extensions to Valuation Methodologies section). THE MATHEMATICS OF A SIMPLE SIMULATION
To give the reader a sense of how the occurrence of default can be modeled, we mathematically describe a simplified model of default for a single asset. Suppose that, conditional on a bond surviving to a time t in the future (where time is measured in years), the probability of default over the next short time period Δt is hΔt, where h is constant. Under this assumption, the time to default follows an exponential distribution; i.e., the probability of the bond defaulting within the next t years is 1 – e-ht. For example, if h=3%, the probability of the bond defaulting within the next two years is 1 – e-.03x2, or 5.82%. The parameter h is known as the hazard rate. exhibit 21
100%
Exponential Distribution: Cumulative Probability of Default
90% 80% Probability
70% 60% 50% 40% 30% 20% 10% 0% 0
10
20
30
40
50
Number of Years
Source: Morgan Stanley
Under this model, the probability of the bond surviving for the next t years is e − ht . Therefore, if we generated a uniform random variable U between 0 and 1, the relation t=-log(U)/h would give us the default time for the bond. For example, using h=3% as above, if in our first path U=0.757, the corresponding time until default would be –log(0.757)/.03 = 9.28 years. One aspect of this simplified simulation that is not realistic is that we have assumed that the hazard rate h is a constant. The hazard rate should be a stochastic function of time calibrated to current market levels. Moreover, to price a CDO tranche using this methodology, we would have one default process for each bond, and would have to combine them so as to account for the fact that defaults among bonds are correlated.
Extensions to the Valuation Methodologies There are many natural extensions to the valuation methodologies that have been described in this chapter. In this section, we identify and summarize some of the extensions that are practiced in the marketplace.
Please see additional important disclosures at the end of this report.
105
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 7
MOODY’S DOUBLE BINOMIAL METHOD
Moody’s applies the Double Binomial Method to CDOs that have collateral from two distinct pools (e.g., high yield bonds and emerging markets debt).3 This technique allows each asset class to have a distinct default probability and diversity score. The two pools are assumed to be uncorrelated. DISTRESSED CDO NOTES AND EQUITY: OPTION VALUE
Given the deep level of subordination that exists in many CDO structures, the marketplace has seen many subordinated tranches lose substantial value to a point where they trade like distressed securities. A distressed CDO does not imply that the entire CDO structure is in financial distress, it implies that the subordinated note of concern is in distress given its position in the cash flow waterfall. For distressed securities, the collateral has experienced par losses (defaults) to a point where the likelihood of receiving future interest and principal payments for a given subordinated note or equity is small. It is difficult to value distressed CDO notes and equity using the valuation techniques we have described in this chapter. Any multi-path approach will be particularly sensitive to the underlying assumptions. Yet these distressed securities do have some value to those investors who are willing to take the risk, similar to out of the money options. In fact, a distressed CDO tranche could be valued as a series of out of the money options, with each potential cash flow modeled as a separate option. The value of the security is the sum of the option value series. From an investor’s perspective, even if one cash flow (coupon payment) is received, the investment could have a positive return if the price of the security is low enough. Consider a hypothetical equity tranche for a CBO. The graph in Exhibit 22 shows annual default rates versus yield for three different prices. At a price of 30, the equity has a yield of 7.4% at a future default rate of 0%, which is the best-case scenario for this residual tranche. However, the yield quickly falls to a value of 0% if annual default rates climb to 2.25%. At a price of 10, the tranche yields nearly 25% for a 0% default rate. The tranche yields about 5% if annual default rates climb to 5%.
3
The Double Binomial Method and its Application to a Special Case of CBO Structures, Moody’s Investor Services, March 1998.
106
exhibit 22
30
Distressed Equity: Yield Sensitivity
Price=10
Price=20
Price=30
Yield
20 10 0 -10 -20 0.0%
0.5%
1.0%
1.5%
2.0%
2.5%
3.0%
3.5%
4.0%
4.5%
Default Rate
Source: Morgan Stanley
NON-BINOMIAL DEFAULT DISTRIBUTIONS AND DEFAULT CORRELATION
The re-rating methodology described earlier in this chapter is based on two generalizations about the default behavior of assets. First, the collateral is generalized to be an N-asset portfolio (N = diversity score) where each asset has the same probability of default (note that the double-binomial Moody’s approach introduces two independent collateral pools). Second, the default correlation among the N assets is assumed to be zero (i.e., their default behavior is independent). This general treatment of the collateral leads to a computationally simple approach to calculating the expected loss of a CDO tranche. However, a natural question to ask is: how different would our results be if we assumed that each credit had a unique probability of default and that the default correlation of assets were non-zero? In practice, computing an expected loss for a tranche in this manner is both data intensive and computationally complex, but it is practiced in the industry. There is commercially available software from at least one vendor (KMV L.L.C.) that provides default and correlation data and supports portfolio loss calculations based on this data. The KMV approach 4 to modeling default risk is based on credit-specific expected default frequencies (EDFs) and a model for default correlation. The EDFs are calculated for each modeled credit and are based on the notion that an issuer will default if the market value of its assets falls below the market value of liabilities. The KMV model for default correlation is based the notion of a joint probability of default for two credits. KMV computes a default covariance matrix using a factor model that is based on diversifiable risk (such as those related to the industry an issuers belongs to) and systematic or non-diversifiable risk (such as those related to a country’s economy and business cycle).
4
“Portfolio Management of Default Risk,” KMV LLC, 31 May, 2001.
Please see additional important disclosures at the end of this report.
107
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 7
exhibit 23
40%
Binomial Distribution
Normal Distribution
Skewed Distribution
35% 30% Probability
Comparing Distributions With Identical Mean and Variance
25% 20% 15% 10% 5% 0% 0
5
10
15
20
Number of Defaults
Source: Morgan Stanley
The resulting loss distributions from the KMV approach are in practice much more skewed than the normal distribution or the binomial distributions we described in the ratings methodology section of this chapter (see Exhibit 23). The skewed behavior comes from positive default correlation of issuers and industries. What does a more skewed distribution imply? KMV’s explanation is that: •
Most of the time, the loss is less than the average loss.
•
The probability of large losses is higher with the skewed distribution (i.e., the skewed distribution has a fat tail). In a normal distribution, the chance of a 4 or greater standard deviation event occurring is very small. In the skewed distribution, it is much higher.
VALUING THE EQUITY OPTION
As we discussed in the market value approach, a change in the market value of the underlying collateral pool is useful in valuing cash flow CDOs in several ways. Related to the market value of underlying assets, there is one feature of CDO equity that has been overlooked so far in this chapter: the value of the option equity holders have to call the entire structure. This call option allows equity holders to delever the structure by forcing a call on all of the CDO’s outstanding debt. Any remaining proceeds after the notes are retired (and after fees are paid) go to equity investors. Equity holders may be motivated to call the structure for a variety of reasons including: •
Capturing market value gains.
•
Limiting losses if the structure could lose more value.
•
Reinvesting capital if equity returns look unattractive on a forward looking basis.
In theory, equity holders will exercise this option when the net present value of resulting proceeds (of calling the structure) is greater the net present value of the future (projected) cash flows. In practice, equity investors would consider the market risks of selling the (potentially illiquid) collateral as well. In this chapter, we have not tried to model the value of this option, but if we were to, the model would be based on the asset price distribution of the underlying collateral (or a
108
proxy to the underlying collateral such as a market index). In past research, we have discussed some of the technical conditions for calling high yield collateral CDO structures and concluded that calls are most likely after a CDO's reinvestment period ends. For the burgeoning class of CDOs backed by higher quality collateral such as investment grade corporates and structured finance, the value of the call option is likely more significant for several reasons. First, the collateral is more liquid and therefore easier to sell in the event that equity holders decide to retire the structure. Second, it is simpler to project the future cash flows of the equity, given the narrower distribution of defaults for the collateral.
Summarizing the Valuation Methodologies In this chapter, we have presented three basic methodologies for secondary market CDO note and equity valuation, namely the re-rating methodology, the market value methodology, and the cash flow methodology. We have presented variations and extensions for these methodologies as well, leaving CDO investors with a broad choice of valuation techniques. We summarize each of these variants in the table in Exhibit 24. In our view, all three basic methodologies, and all of their variants have their valid uses, but they differ in the degrees of computational complexity and required market savvy. THE NEED FOR STANDARDIZATION
Which methodology should investors use? Investors and traders may never all agree on the best approach to value CDOs. In our view, a standardized approach to valuing CDOs would give investors a common language to speak and at the same time improve liquidity. As an analogy, in the mortgage-backed securities market, standardized prepayment assumptions and generalized collateral pools simplified the market tremendously, with sophisticated investors and traders relying on proprietary prepayment models to perform further valuations. One of the goals of this chapter is to encourage more dialogue among industry participants on the topic of standardization. COMPUTATIONAL COMPLEXITY VERSUS MARKET SAVVY
In the absence of standardization (or in addition to standardization when it exists), there are differences in the popularity of the various methodologies. We find the market value and re-rating approaches are currently more popular than cash flow approaches, due largely to their computational simplicity. In the near term, we believe computational complexity will be a very important factor in determining the popularity of valuation methodologies. There are only a handful of commercially available or investment bank-provided CDO analytics tools today, so investors are naturally favoring the computationally simpler approaches that necessarily require a good deal of market savvy. The analytics Morgan Stanley provided are among the most sophisticated, allowing investors to value CDOs using the multi-path cash flow methodology. We expect the scale to tip in the other direction over time, with investors having many options to value CDOs both from commercial analytics providers and investment banks. In the interim, the choice of methodology will be a function of the tools available to an investor as well as the investor’s familiarity with CDO structures and the underlying collateral markets.
Please see additional important disclosures at the end of this report.
109
110 Description • Generalize collateral by industry group • Industry groups are uncorrelated • Assume one default rate (two for double binomial) • Calculate expected loss under binomial distribution • Infer a credit rating, determine value by comparison
For low default rate Weak model support assets, equity option can drive value of the equity tranche
• Value the option of equity holders to call entire structure
Equity Option Value
Structural changes may not ultimately influence value and can be subjective
Forecast CDO structural changes
• Use market value as an indicator of future CDO structural change such as ratings action, defaults, potential manager transactions • Forecast structural changes and value using Re-rating or Cash Flow methodology
Difficult to assign market value changes to individual tranches
Indicator of Structural CDO Change
Market Value Market Moves in Spreads and Interest Rates
Simple asset and liability equivalence relationship
Lots of market data
Disadvantages Very general collateral and default assumptions No un-rated equity valuation
• Observe market value change of collateral • For small changes, adjust subordinated note and equity values through market comparisons • For large changes, adjust all note values and fundamental equity assumptions (discount rate)
Default and correlation model with credit specific information
Advantages Simplest approach
CDO Valuation Methodologies
High
Medium
Medium
High
High
High
High
High
Required Computational Market Complexity Savvy Low High
Low
High
High
Low
Popularity Among CDO Investors High
chapter 7
Non-Binomial • Use credit-specific, model-generated default rates and correlation Default Distribution (Can • Calculate expected loss under this unique distribution be variant of Cash • Determine value by comparing expected loss Flow)
Valuation Methodology Variant Binomial Re-rating Distribution
exhibit 24
Structured Credit Insights – Instruments, Valuation and Strategies
Please see additional important disclosures at the end of this report.
111
Fair market approachRequires sophisticated model
Fair value for the series of options
• Calibrate simulator to model arrival time of defaults over a large distribution • Generate cash flows over this distribution, discounting by zero coupon Libor • Average results to determine value • Model tranche like an out-of-the-money option • Determine option value
Simulation
Distressed
Source: Morgan Stanley
Comparables based Difficulty in finding the right comparable with sensitivity testing
• Select a discount rate (yield) by comparing tranche to similar structures in the market • Discount cash flows with this discount rate under a series of input assumptions • Observe price sensitivity to changes in assumptions
Multi Path with Sensitivity Analysis
Weak model support
Advantages Simple benchmark estimate
Disadvantages One path too restrictive, purely theoretical value
CDO Valuation Methodologies
Description • Adjust tranche cash flows by default expectations • Discount cash flows by zero coupon Libor to determine value
Valuation Methodology Variant One Path Cash Flow
exhibit 24 (cont.)
Medium
High
Medium
High
Low
Medium
Required Computational Market Savvy Complexity Medium High
Low
Low
Medium
Popularity Among CDO Investors Low
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 8
Portfolio Sampling Risks
June 1, 2002
Primary Analyst: Sivan Mahadevan David Schwartz A common misconception in the analysis of CDOs is the assumption that the underlying collateral pool performs just like the “market” or just like an “index” for a given level or risk. Such generalizations make historical stress analyses easy, as all one has to do is observe the historical performance of the market or index. However, CDO collateral pools tend to have fewer securities than any broad measure of the market, and as such, there is a significant risk that the collateral pool behaves differently than the market, even if aggregate risk measures such as ratings are similar. Such portfolio sampling error is not necessarily a bad thing. In fact, investors who choose a collateral manager to actively manage a portfolio are effectively paying them a fee to deviate from the market with the hope that the deviation is beneficial to them. However, there is always the risk that the portfolio deviates from the market in a negative manner as well. We address this notion of portfolio sampling risk in this chapter. TRACKING ERROR: TOTAL RETURN VERSUS DEFAULT RATES
There are many ways to measure the “risk” of a sample portfolio relative to a broad measure of the market. Index “tracking error” is perhaps the industry’s most popular approach and involves observing the distribution of periodic total return differences between a portfolio (or portfolio strategy) and an index. The standard deviation of this difference is one measure of this tracking error. For a cash flow CDO, rather than periodic return, the default rate is the variable whose distribution impacts the performance calculation. The risk that a particular tranche of a CDO misses its targeted return is directly related to the probability that the default rate exceeds a certain threshold (the probability of note impairment). From the mezzanine note investor’s perspective, it is important to measure is the sensitivity of the probability of impairment to the size of the underlying collateral portfolio. To measure this sensitivity, we first need to generate the distribution of defaults for a given portfolio strategy. The traditional approach in the CDO market for doing this is to model portfolio default behavior using a binomial distribution. Our model generates what we consider to be a more realistic default distribution, by assuming positive default correlation between credits. MEZZANINE NOTE IMPAIRMENT: SENSITIVITY TO PORTFOLIO SIZE
In Exhibit 1 we show the probability of impairment as a function of the size of a portfolio, for various mean default rates. The X-axis of the graph is the number of issuers in the portfolio. The Y-axis is the probability that the underlying collateral pool experiences enough defaults to impact the cash flow of the mezzanine notes. This analysis is based on a Baa2-rated mezzanine tranche of a typical 5-year synthetic investment grade CDO, whose underlying portfolio has an average rating of Baa1/Baa2.
112
It is important to note that adding more and more credits to a portfolio does not reduce the probability of mezzanine note impairment to zero. This is true for two reasons. First, because the underlying credits are assumed to have positive default correlation, some of the risk is systematic and cannot be diversified away. Second, there are only a limited number of credits to choose from. We estimate that the global credit investor has access to approximately 1,200 investment grade credits, through the framework depicted in the Venn diagram in Exhibit 4. To arrive at this estimation, we first considered the universe of credits available in the two largest corporate bond markets (the US dollar and Euro-currency markets, with credits from all over the world). We also considered the credit default swap markets in the US and Europe. While this methodology does not result in the comprehensive universe of issuers available, it is a reasonable measure of total available credit exposure. Adding corporate credits that trade only in selected domestic markets (such as the United Kingdom, Japan, and Australia) may add some credits to the total (say an additional 200 to 300) but accessing these credits can be difficult, even for the most global of investors. In our view, the 1,200 estimation should be considered the upper bound for the number of credits in a global investment grade credit portfolio that could realistically be included in a CDO. In light of the fact that some of the risk that a mezzanine note becomes impaired cannot be diversified away, we find it useful to consider the “diversifiable risk” of impairment, shown in Exhibit 2. For this calculation, we take the probability of impairment and subtract the non-diversifiable risk, as measured by the probability of impairment for a 1,200 credit portfolio. exhibit 1
Probability of Mezzanine Note Impairment vs. Portfolio Size
% 10
expected default rate = 3.5% expected default rate = 2.5% expected default rate = 1.5%
8 6 4 2 0 0
200
400
600
800
1000
1200
Portfolio Size
Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
113
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 8
exhibit 2
Diversifiable Risk: Probability vs. Portfolio Size
% expected default rate = 3.5% expected default rate = 2.5% expected default rate = 1.5%
6
4
2
0 0
200
400
600
800
1000
1200
Portfolio Size
Source: Morgan Stanley
exhibit 3 5-Year Cumulative Default Rate 1.5%
Table of Inflection Points and Sweet Spots Inflection Point 150
Sweet Spot 300
2.5%
190
400
3.5%
210
500
Source: Morgan Stanley
For a given mean default rate, we identify two points on this curve that we think are very important from the perspective of mezzanine note investors. The first is the “inflection point”. 1 We define the inflection point as the point at which a sufficient amount of the diversifiable risk has been removed. The second is the “sweet spot.” We define the sweet spot as the point at which the marginal benefit of additional credit exposure is sufficiently small. At a 1.5% default rate (which corresponds to the Moody’s idealized 5-year cumulative probability of default for a Baa2-rated portfolio), the inflection point is approximately 150 credits while the sweet spot is approximately 300 credits. At a 2.5% default rate, the inflection point is approximately 190 credits and the sweet spot is approximately 400 credits. At a 3.5% default rate, the inflection point is approximately 210 credits with a sweet spot of 500 credits.
1
We are not referring to the mathematical definition; i.e., where the second derivative of the graph changes sign. Rather, we have in mind the colloquial definition of “a turning point, a point which marks significant change.”
114
exhibit 4
Global Cash and Credit Default Swap Markets
US Dollar Cash Corporate Market 800 Credits
US CDS 325 Credits EUR Cash Corporate Market 250 Credits
European CDS 325 Credits
Source: Morgan Stanley
THE STATIC VERSUS MANAGED DEBATE
Any analysis of portfolio sampling risk within CDOs is usually at the heart of a broader discussion: the debate over investing in static deals versus hiring a manager to actively manage the collateral. Managers are generally favored in markets where credit skills and access to assets are limited, such as high yield bonds, leveraged loans, and emerging markets. However, in markets where such skills are broadly available and where access to assets is not as difficult, CDO investors truly have a choice between managed or static transactions. There are clear benefits and drawbacks for each approach. With a manager, investors hope that there is significant portfolio sampling error that leads to outperformance relative to the market. However, there is a cost for management, so any managed strategy has to be credible (in the minds of the investors) and should leave the manager with enough room to deviate from the market to earn the “alpha.” In contrast, in a static transaction, investors are hoping that either initial credit selection will be adequate (with respect to default risk) for the life of the transaction, or that the collateral pool is large enough that the risk of tranche impairment is sufficiently low. The analysis provided in this chapter should aid investors in understanding the potential risks they have as mezzanine note investors, given the wide differences in CDO portfolio size in the marketplace. CONCLUSION
As we have highlighted in this chapter, portfolio size is an important consideration for mezzanine note investors. If a CDO’s underlying portfolio is larger than the inflection point (for a given default rate assumption), much of the diversifiable default risk can be significantly reduced. If the portfolio is larger than the “sweet spot” size, the marginal benefit of adding additional credits is small. It is important to note that such risks may be desirable in cases where investors intend to take name-specific credit risk in either static or managed transactions.
Please see additional important disclosures at the end of this report.
115
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 9
Streetwise Correlation
June 13, 2003
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Anisha Ambardar
Correlation trading is the new new thing in credit markets. Over the past two years, CDO investors have become painfully aware of the correlation risks in their structures. As a result, investors are now focused on understanding the correlation of credits in portfolios backing synthetic and cash structures, although there is still healthy debate on what this actually means. As correlation relationships can be puzzling, our goal in this chapter is to help investors gain some intuition on the topic. We have been writing about correlation risks in tranched investments for at least 18 months now, but a lot has occurred in the past 6 months. •
The Street has standardized a technique to price baskets, given correlation assumptions
•
A market for first-to-default baskets has emerged, and we can observe implied correlation from these transactions
•
A “trading” market for tranches of large baskets is developing, based on Synthetic TRACERSSM, and we can observe implied correlation from this market as well
STREET SPEAKING THE SAME LANGUAGE
We refer readers to previous research for a more detailed description of default correlation, and for a discussion on how to estimate it using structural models of default (Merton or Moody’s KMV-like). Over the past several months, the Street has standardized a quantitative approach to pricing baskets. Without going into the details, we simply describe it as a way of taking a collection of independent default distributions (for each company) and combining them into correlated distributions for all of the companies in a basket. From this pricing process, an implied correlation can be observed. This pricing model can be thought of as the Black-Scholes for correlation products, although many have argued that other approaches should not be dismissed, including those that use a real correlation “matrix” instead of an average. Nevertheless, the beauty of this approach is that the Street now speaks a single language, and market participants can observe implied correlation, much like implied volatility in the options markets. RULES OF THUMB
There is still a lot of correlation confusion among market participants. We argue that the rules of thumb are fairly simple.
116
•
In “subordinate” tranches, risk and spread decrease as correlation rises
•
In “senior” tranches, risk and spread increase as correlation rises
Investors in “subordinate” tranches are long correlation (spreads tighten when correlation rises). Investors in “senior” tranches are short correlation (spreads widen when correlation rises). In Exhibit 1, we illustrate these relationships for a simple fivename basket. exhibit 1
Spread
Correlation Relationships – Small Baskets
First To Default
10%
20%
30%
40%
50%
60%
Second-Fifth To Default
70%
80%
90%
100%
Correlation
Source: Morgan Stanley
THE “SUBORDINATE” VS. “SENIOR” DIVIDE
As we graduate from small baskets to large baskets, two things happen. First, the range of possible (or implied) correlation values narrows as large baskets are less sensitive to the idiosyncratic behavior of a small number of companies. Second, the tranche sizes that trade are typically much thinner than in small baskets, which makes their interrelationship very interesting from a correlation perspective. As an example, dealers are now making markets in five tranches of synthetic TRACERSSM, all within the bottom 30% of the capital structure. By comparison, in a five-name first-to-default basket, one can think of the implicit “tranche” size as being 0%-20%. Given the number of tranches in the first 30% of the capital structure, we can think of each tranche as being somewhere on the spectrum from “subordinate” to “senior.” In Exhibit 2, we focus on the sensitivity of four tranches (ranging from 3% to 15% attachment points) of synthetic TRACERSSM and make a few observations. •
The 3%-6% tranche has a downward sloping spread/correlation relationship (long correlation), suggesting that it is a “subordinate” tranche.
•
The 6%-10% tranche has an odd convex shape; its spread first widens with rising correlation (like a senior tranche) but then tightens (like a subordinate tranche).
•
The 10%-15% tranche is mostly upward sloping, with a curve that changes shape only at a high correlation, implying it is in the “senior” category.
•
The 15%-30% tranche appears to be strictly upward sloping, suggesting it is a pure “senior” tranche under all reasonable correlation assumptions.
Please see additional important disclosures at the end of this report.
117
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 9
One point worth noting is the interaction between spreads and correlation. The average spread of the TRACERSSM portfolio in this analysis is approximately 70 basis points. If the spreads of the underlying portfolio widened dramatically, we can expect the 6%10% tranche to behave more and more like the current 3%-6% tranche with respect to changes in correlation. exhibit 2
Spread 3-6%
Correlation Sensitivity – Tranched TRACERSSM
10%
20%
30%
40%
6 - 10 %
50%
60%
10 - 15 %
70%
15 - 30%
80%
90%
Correlation
Source: Morgan Stanley
INTUITION
In a nutshell, our rules of thumb still apply for large baskets, but the definition of a “subordinate” tranche depends on the initial correlation assumption and portfolio spread levels. The intuition is that correlation measures how risk is distributed among the tranches. Low correlation implies that risk is concentrated in the most subordinate tranches. High correlation implies that risk is distributed among all tranches. WHERE DOES CORRELATION TRADE? SMALL BASKETS
Based on market data from our benchmark first-to-default baskets, we can easily observe where correlation trades for small baskets. As an example, for the diversified industrials basket, mid-market implied correlation has been approximately 45%, while long-run average pair-wise equity correlation of the five companies is 31%. For the slightly more concentrated basic industrial basket, the average equity correlation is 42% while the basket trades at approximately 50%. Although not directly comparable (one could write a PhD thesis on why this is so), we do not consider this disconnection between equity and market-implied correlation to be significant. The concentrated and idiosyncratic nature of smaller baskets makes them less risky for sellers of protection on a first-to-default basis, which is why correlation trades higher.
118
WHERE DOES CORRELATION TRADE? LARGE BASKETS
In Exhibit 3, we show the new correlation information from the large basket market (100-name synthetic TRACERSSM). We make two important observations. First, implied correlation is generally lower than historical equity correlation. We are seeing the opposite behavior of the small basket market, namely the larger basket size is putting downward pressure on correlation (the diversification impact). Second, in an arbitrage-free world, a given capital structure should trade at the same correlation, but we are clearly observing significant correlation differences among tranches. These differences demonstrate both the nascent nature of the market and the difficulty to arbitrage (transaction costs). exhibit 3
TRACERS
TRACERS
SM
SM
Correlation Observations – Large Baskets Bid-Offer Spread (bp)
Implied Correlation
0%-3%
43/51 & 500 pa
24%
3%-6%
540/610
33%
7%-10%
145/185
19%
10%-15%
75/95
27%
15%-30%
14/23
31%
Tranches
Equity Correlation 1 Year Daily
33%
10 Year Monthly
38%
SM
Note: TRACERS equity correlation are historical observed values. 0%-3% tranche trades with points upfront. Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
119
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 10
Correlation Conversations, Convexity Ideas
August 1, 2003
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Anisha Ambardar
With lower levels of volatility in the credit markets and a fairly stable spread environment recently, correlation markets have further matured. We can point to increased participation in both the first-to-default basket and the larger single-tranche markets, as investors simultaneously reach for yield and discover tools to implement market views, either on the long or short side. In our recent travels, we have discussed the state of the correlation market with a number of institutions and have discovered quite a few sophisticated, cutting-edge correlation investors. That group would have been significantly smaller a year ago. More importantly, there is a much larger class of investors who are just stepping into this market, and the resource allocation decisions they make over the next few months will be a key development for the market going forward. We focus this chapter on providing some insight into the recent correlation moves within the market’s benchmark basket products and recommend a leveraged strategy for getting short credit that can take advantage of an interesting convexity relationship. CORRELATION AND MINEFIELDS
Gaining intuition on correlation remains a challenge for many market participants. Our European strategy team has had success in using the following analogy: Consider a minefield, where an investor’s goal is to make it across without blowing up. A subordinate tranche investor will blow up after he hits, say, two mines during his journey, while a senior tranche investor is safe until she hits ten mines. If the enemy places 20 mines throughout the field, where would the subordinate tranche investor like them to be placed? Clearly he would like them to be clustered in one small area, as this reduces the chance that he will hit one as he walks across. If they are dispersed evenly across the field, that is worse for him as the chance of him hitting one on any given path is actually higher. Remember, he is not concerned about hitting more than two. The senior note investor will think of it differently. If the mines are dispersed across the field, chances are that she will hit a few but the chance of hitting ten is unlikely. If they are clustered in one area, she may make it through safely on several paths, but if she hits the cluster, she will most likely blow up. We can think of the correlation in this framework. A cluster of mines is high correlation (which the subordinate tranche investor likes), while mines dispersed more evenly throughout the field is low correlation (which the senior note investor prefers). CORRELATION INSIGHTS
Since tranched Dow Jones TRAC-XSM (formerly called synthetic TRACERS) started trading in late May, we have collected some very interesting data on relative correlation moves of the various tranches during a period when the underlying
120
portfolio has traded in a narrow range. The correlation disparities we discussed in Chapter 9 have actually grown wider as flows have been tilted toward sellers of protection, particularly in “senior” tranches (see Exhibit 1). exhibit 1
Diverging Implied Correlation – Tranched TRAC-X
45% 40%
0-3% Mid Imp Correl
3-6% Mid Imp Correl
7-10% Mid Imp Correl
10-15% Mid Imp Correl
15-30% Mid Imp Correl
35% 30% 25% 20% 15% 10% 05-Jun-03
21-Jun-03
07-Jul-03
23-Jul-03
Source: Morgan Stanley
In early June, most tranches traded in a 24-33% implied correlation range, but today implied correlation values are significantly lower for “senior” tranches while being much higher for the “mezzanine” tranche (3-6%). These correlation moves are almost entirely attributed to flows. Therefore, a reach for yield among investors has driven both the 7-10% and 10-15% tranches tighter, moving correlation lower. WHO IS RIGHT?
We illustrate in Exhibit 2 today’s implied mid-correlation for the various tranches (the squares) on curves that show the respective spread and correlation relationships. The dashed vertical line is at a 20% correlation, corresponding to where the 0-3% (equity) tranche is currently trading. exhibit 2
Who Is Right? – Correlation Divergence
“Fair” Spread 700
3-6%
7-10%
10-15%
15-30%
600 500 400 300 200 100 0 10%
20%
30%
40%
50% 60% Correlation
70%
80%
90%
Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
121
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 10
So what conclusions can we draw from this increasing correlation discrepancy? We have no strong recommendation at this point, but do have some insights. Long-run historical equity correlation is in the 30% range for the TRAC-X portfolio. The 0-3% tranche is trading far from that, but also has a large “error” term, given the tranche’s sensitivity to correlation and wider bid-offer compared to other tranches. Other correlation values are fairly scattered around this 20% level, so it is hard to draw any conclusions from this analysis alone, which leads us to focus on trading strategies as a way to think about opportunities. AVOID THE ARBITRAGE TEMPTATION
Although very tempting, given that bid-offer spreads have compressed, we do not recommend capital structure arbitrage trades in tranched TRAC-X. We do not see any reason why correlation values should converge in the near-term. The market is still new, pure correlation traders do not exist in large numbers and many investors are driven by yield and rating as opposed to “fair” pricing from derivative pricing models. We further add that, although the Street has standardized on a common pricing approach, there are numerous end investors who take a very different approach to modeling correlation, based on techniques that are more closely linked to the equity markets. For all these reasons, the correlation discrepancies are explainable and could remain in the near term. Tranche relative value trades based on expected spread changes or flows might be a better approach to exploiting these relationships, which we discuss next. GETTING LONG CONVEXITY FROM THE SHORT SIDE
The trade that stands out as most appealing, in our opinion, comes from the view that we find a lot of comfort in getting short tight-trading names in today’s markets (see “Finding Comfort on the Short Side,” July 11, 2003). In particular, implementing a leveraged short credit view by buying protection on the 10-15% tranche appears attractively priced, considering spread levels and the convexity benefit. exhibit 3
Convexity in Senior Notes for Short Sellers (Carry Neutral P/L as % of Notional)
20% 3-6% 7-10% 10-15% 15-30%
15%
TRAC-X
10%
5%
0%
-5% 40
50
60
70
80
90
100
110
120
130
140
150
TRAC-X Spread at Trade Exit
Source: Morgan Stanley
With TRAC-X protection trading at about 72 bp on the offered side, investors can buy protection on the 10-15% tranche of TRAC-X for 62 bp (implied correlation of 20%). The 10-15% tranche will behave like a levered investment as spreads rise. This leverage will continue to increase as spreads widen (see Exhibit 3). As an example, consider the tranche’s theoretical value as TRAC-X rallied over 100 bp over the past
122
10 months (see Exhibit 4). The 10-15% tranche would have experienced a 400 bp move, representing on average approximately 5x price leverage. While we do not expect 100 bp of widening any time soon, the historical example simply illustrates that this tranche is indeed a levered play on credit on the downside. Clearly, this type of leverage is interesting to investors who want to implement a short view, but the key benefit to this trade recommendation is the positive convexity the protection buyer gets when spreads tighten. The protection buyer in a 10-15% tranche will certainly feel pain as the market rallies, but the losses will be less severe than the gains on the other side (see Exhibit 3). Why is this so? In the most basic case, the tranche’s value is floored at zero, so a big relative rally in spreads has limited downside for the buyer of protection. Furthermore, given this floor, the tranche’s price movement may be much more muted as spreads rally. Our model tells us that a 30 bp tightening of TRAC-X (to a 40 bp level) would result in a “fair spread” of 14 bp on this tranche. If the tranche did trade at that level, it would represent less than 2x leverage, which we would argue is pretty good convexity for the buyer of protection. However, this is uncharted territory for TRAC-X, and we are not certain that the tranche would actually trade that tight, given practical market issues. We can point to the European credit markets to gain a bit of insight on valuation. TRAC-X Europe (100 names) trades in the 40 bp range today. Pricing in tranches of TRAC-X suggest that a 10-15% tranche would trade at higher levels than the “fair” value, possibly in the 25-35 bp range. This results in even a better convexity story for the short-viewed investor who uses this tranche to implement his or her view. exhibit 4
Tranche Fair Value Historically – Illustrating Leverage
1,200
1,000
Tracers
6 - 10%
10 - 15%
800
600
400
200
0 Oct-02
Dec-02
Jan-03
Mar-03
Apr-03
Jun-03
Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
123
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 11
How Many Sides in a CDO Squared?
January 23, 2004
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Anisha Ambardar
The market for CDO transactions backed purely by other CDOs never really blossomed, for a variety of reasons. Analyzing portfolios of CDO tranches using stateof-the-art models still remains a difficult and computationally intensive task. While such CDO squared transactions (as they are known colloquially) can benefit from a lack of first loss exposure in the underlying portfolio (thereby making them more immune to the dreaded idiosyncratic risk phenomenon), investors fear concentration risk across deals that is hard to either control or monitor. Yet, CDO securities have become an increasingly common part of the collateral pool for both structured finance and synthetic CDOs. Secondary CDO “buckets” were typically small in the past, but today’s spread levels in investment grade and ABS markets have pushed many transactions to hunt for yield in other ways. The newly established CDO secondary market has been the answer for many. In CDOs backed by asset-backed and commercial real estate securities, structures now allow secondary CDO exposure – up to 35% in some cases. “Synthetic” buckets can allow even more room for CDO exposure. Similarly, in synthetic CDOs, single-name credit default swaps are also combined with other synthetic CDO tranches to form the reference portfolio. What are the risks in these structures, and for whom are they best suited? We attempt to shed some light on the multi-dimensional puzzle in this chapter. THE BASIC STRUCTURE AND MARKET
Despite all the fanfare several years ago, the process of resecuritizing portfolios of outstanding CDOs into a new CDO transaction remains a small part of the overall market. In 2003, there was only approximately $1 billion of CDO squared issuance, a small fraction of the $65 billion new issue market. Yet, we estimate that several billion dollars of secondary CDO paper found their way into both structured finance CDOs (where total new issuance was over $20 billion) and synthetic structures. In fact, such vehicles made up 15% of the CDO buyers in the secondary market in 2003, by our estimates. These structured finance and synthetic CDOs are really hybrids (see Exhibit 1).
124
exhibit 1
CDO Hybrid Structure
The Anatomy of a Hybrid CDO Structure Senior Notes
Collateral
Corporate Credit and/or Asset Backed Securities Corporate Credit and/or Asset Backed Securities
Mezzanine Notes
Equity
Secondary CDO Tranches
Corporate Credit and/or Asset Backed Securities Corporate Credit and/or Asset Backed Securities Etc.
Source: Morgan Stanley
WHAT’S THE RATIONALE FOR THIS TRADE?
At the 10,000 foot level, the benefits of including secondary CDO paper in primary CDO transactions are fairly obvious. First, subordinate CDO tranches tend to be “cheap for the rating” because ratings are generally based on default risk, ignoring both the positive and negative mark-to-market implications of the leverage and convexity. Second, CDO tranches have less idiosyncratic risk than any single-name investment, given a lack of first loss exposure. In addition, the advocates will claim that even stressing historical default rates will still result in most structures avoiding losses. Third, CDO tranches can be a good means of diversifying into other types of collateral. There is a lot of merit to the above argument if default risk (and associated ratings migration) is the only concern an investor has. Yet the leverage and convexity that is inherent in CDO tranches makes CDO squared (or hybrid) structures potentially more volatile than plain vanilla structures. Furthermore, what is missing from this argument is a dose of reality. There are many investors who do care about mark-to-market implications. Reduced idiosyncratic risk, while true in theory, becomes diluted when the various secondary CDO tranches reference portfolios that are “similar” (i.e., that have overlapping credit risk). Finally, we reiterate that secondary CDO tranches are not 100% of the collateral in today’s typical transactions, so the relationship (correlation) between these secondary tranches and other credit risk is important. SQUARES OR CUBES – HOW MANY SIDES TO THIS PUZZLE?
When analyzing CDO structures that include secondary CDO paper, there are three dimensions that warrant close attention, in our view. The first dimension is the accounting regime of the investor. As we mentioned above, the leverage and convexity of the tranches can lead to more volatile mark-to-market valuations, and investors should be comfortable with their exposure (if any) to these market swings. The second dimension is the seniority of the secondary CDO tranches included in the collateral, while the third dimension is the seniority of the CDO squared tranche in question, both of which again can have substantial valuation implications.
Please see additional important disclosures at the end of this report.
125
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 11
exhibit 2
Collateral
The CDO Squared Puzzle – Senior or Mezzanine Exposure to Senior or Mezzanine Collateral?
Senior
Mezzanine
•Little idiosyncratic risk
•Ratings stability
In between the two extremes
•Negative convexity (m-gamma) •Impact of overlapping credits is small
•Leverage and yield Mezzanine
CDO Squared Structure
Senior
•Very little default risk
•Lots of idiosyncratic risk (i-gamma) In between the two extremes
•Convexity is positive (m-gamma) •Overlapping credits can be disastrous
Source: Morgan Stanley
We highlight some benefits and risks in the second and third dimensions, from the perspective of an investor going long the tranches, in Exhibit 2. Of the four possible “squares” in the diagram, the two extremes are senior exposure to a portfolio of senior tranches (where idiosyncratic and ultimate default risk are low but convexity is negative) and mezzanine exposure to a portfolio of mezzanine tranches (where leverage and idiosyncratic risk are high, but yield and positive convexity are the benefits). As for the other squares, the benefits and risks get mixed a bit. For example, mezzanine exposure to a portfolio of senior tranches still has low idiosyncratic risk but is more levered and higher yielding than senior exposure to the same portfolio. The risk of overlapping reference credits is largest when mezzanine tranches are included in the collateral pool. QUANTIFYING THE MTM IMPACT
The above “square” illustrates some of the pros and cons of CDO squared structures, but we have yet to shed any light on the first dimension: mark-to-market implications. Doing so requires making many assumptions, but here is one way to build up a framework. Consider the following worst-case scenario: a CDO squared structure where all of the underlying CDO tranches are identical, meaning they have the same attachment and detachment points and reference identical underlying portfolios (say Dow Jones TRAC-X NA II). Such a structure is an unrealistic trade, but it gives us the basis for analyzing some of the risks.
126
As an example, if the Dow Jones TRAC-X portfolio moves 25 bp wider, a 7-10% tranche on this portfolio would widen 116 bp (based on the model) and a 10-15% tranche would widen 61 bp (see Exhibit 3). Now a 7-10% CDO squared tranche referencing a portfolio of identical 7-10% Dow Jones TRAC-X tranches would widen 906 bp, which we consider a worst-case scenario, owing to no diversification. But more realistically, if the CDO had 10% exposure to 7-10% tranches on TRAC-X-like portfolios, then it would widen by less than 91 bp, depending on the number of credits in the portfolio and the exact structure. This assumes that the remaining collateral portfolio is unchanged. The benefits of seniority are clear if we consider the same situation with 10-15%-type underlying tranches. The moves in the CDO tranches would be at most (but likely less for diversified structures) 35 bp and 18 bp, respectively, for 7-10% and 10-15% exposure to 10-15% tranches. exhibit 3
Underlying Tranches (bp)
25 bp Move Wider in Dow Jones TRAC-X NA II Results in… 100% Exposure (No Diversification) 7-10% 10-15% 116 61
Upper Bound on 10% Exposure (More Typical) 7-10% 10-15%
CDO Squared Tranches 7-10% (bp)
906
350
91
35
10-15% (bp)
496
182
50
18
Source: Morgan Stanley
OUR VIEW, IN A NUTSHELL
CDO squared structures are complicated business, but we have some concluding thoughts for investors who are considering getting long tranches in the large market of hybrid CDOs structures. For investors who are less sensitive to mark-to-market issues, such structures can offer a significant amount of protection from both idiosyncratic and market-wide defaults (depending on the nature of exposure). For investors who are sensitive to mark-to-market issues, it is important to get comfortable with the potential spread moves that we described (on the right hand side of Exhibit 3). In either case, we encourage market participants to form a view on what type of exposure and default protection make sense for the investment. Finally, be careful of overlapping credit risk in the secondary CDO tranches; this is the land mine that may be hard to spot.
Please see additional important disclosures at the end of this report.
127
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 12
Tailoring Baskets – One Size Doesn’t Fit All
April 16, 2004
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Anisha Ambardar
While default swap indices have introduced a good deal of uniformity for the market’s liquid structured credit vehicles, the subtle differences both within and between on-therun and off-the-run opportunities can have important investment implications. One example is in the correlation space, where tranches can have different or even opposite reactions to changes in the number of credits in the underlying baskets. In the early days of the synthetic CDO market, there was a lot of experimentation with basket sizes, with deals ranging from 50 to over 250 credits. In today’s markets, much of the liquidity is centered around 100-name-like portfolios, but customization allows investors to still do what they please. In the cash CDO world, leveraged loan CLOs typically include 50 to 100 credits. Should investors have a preference for larger or smaller portfolios? How much spread should they demand (or give up) to move from a smaller basket to a large one? How sensitive is valuation to different basket sizes? Intuition tells us that long credit investors in senior tranches should prefer larger portfolios (the “diversity is good” argument), while those in subordinate tranches should prefer smaller portfolios (less credits to worry about). The popular correlation models agree with this logic in general, but may not capture market sentiment precisely. In particular, some of the correlation skew that we observe in the marketplace may be explained by views investors hold regarding the randomness associated with default events. Furthermore, the sensitivity of tranche valuation to basket sizes can change under different spread and correlation regimes. SOME INTUITION BEHIND BASKET SIZES
Credit investors focus quite a bit on broad measures of the market such as popular corporate bond indices (800+ issuers) and the rating agency default studies that include an even larger number of companies. As such, comparing manageable-sized portfolios to the market involves a fair amount of sampling risk. Using this perspective, senior tranche investors should prefer larger baskets, because the sampling risk (relative to the market) is smaller. Such investors essentially take comfort in knowing their exposure is “similar” to the market when baskets sizes are large. Furthermore, they reduce exposure to “random” defaults, even if these baskets include more questionable credits. First-loss tranche investors should have the opposite view. They prefer portfolios that will perform better than the broad market, but are very exposed to “random” idiosyncratic events. From their perspective, more credits translate into more headaches, even if the ultimate impact of a single credit event is smaller as the number of credits increases.
128
While the market has a lot of performance experience with baskets of varying sizes, the impact of the credit cycle overshadows these observations. For example, one 250-name static basket issued in 2000 (before the worst of the cycle) has experienced 13 credit events while another 250-name managed deal issued in 2002 (near the end of the cycle’s worst part) experienced none. WHAT IS IT WORTH? WHAT DO THE MODELS SAY?
One way to determine the impact basket sizes have on tranches is to see what the models say, since we do believe the models are grounded in some of the same theory we described above. exhibit 1
Tranche Spread Sensitivity to Basket Sizes
Spread as % of 100-Name Basket
0-3%
2-5%
120%
7-10%
10-15%
For 3-7% tranches, risk decreases as the number of credits in the basket increases
115% 110%
3-7%
The 2-5% tranche risk level remains relatively stable as the number of credits in the basket increases
105% 100% 95% 90%
For first loss tranches, risk increases as the number of credits in the basket increases
85% 80% 50
75
100
125
150
175
200
225
250
Number of Credits in Basket
Source: Morgan Stanley
At today’s spread and correlation levels (62 bp for Dow Jones TRAC-X), the models tell us that the 0-3% tranche rises in spread by about 10% (174 bp) as a basket grows from 50 to 100 credits (see Exhibits 1 and 2). From 100 to 250 names, the rise in spread is only about 6% (107 bp) total. exhibit 2
Tranche 0-3%
How Much Is It Worth? Tranche Spreads for “Similar” Baskets of Different Sizes
50 Names 1,625
100 Names 1,799
125 Names 1,829
250 Names 1,906
2-5%
654
647
645
646
3-7%
409
352
345
320
7-10%
132
117
114
108
10-15%
59
53
51
49
Source: Morgan Stanley
The 2-5% tranche is fairly invariant to basket sizes. The 3-7% tranche is the most sensitive, and has the opposite behavior to the first-loss tranche. Moving from 50 to 100 credits reduces spread by 15% (57 bp), and from 100 to 250 credits is worth about 32 bp. Clearly this junior mezzanine tranche benefits from an increase in the number of
Please see additional important disclosures at the end of this report.
129
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 12
credits. The more senior tranches (7-10%, 10-15%) are shaped similarly to the 3-7% tranche, but are a bit less sensitive. CORRELATION AND SPREAD SENSITIVITY
The above analysis assumes today’s levels for both spread and correlation, but valuations are sensitive to changes in these assumptions as well. As a good example, in a wider spread environment, the 3-7% tranche sensitivity to basket sizes would decrease, with the opposite being true in a narrower spread regime (see Exhibit 3). exhibit 3
3-7% Sensitivity to Basket Sizes Changes Under Different Spread Regimes
Spread as % of 100-Name Basket
140%
Today's Spreads
130%
20 bp Tighter 20 bp Wider
120% 110% 100% 90% 80% 70% 60% 50
75
100
125
150
175
200
225
250
Number of Credits in Basket
Source: Morgan Stanley
WHAT IS IT WORTH? WHAT DO THE MARKETS SAY?
If the markets believed entirely in what the models said, then the implied correlation levels for “similar” risk baskets of different sizes would be identical. Yet, based on some of our observations, implied correlation differences do exist for baskets of varying sizes, although we caution that comparing any two portfolios generally involves differences in credit risk as well. We saw above that risk in first-loss tranches rises as the number of credits increase, according to the models. Investors, however, seem to demand even more premium. As an example, first-to-default baskets trade at high correlations (40-60%) relative to firstloss tranches of larger baskets (20%), implying that sellers of protection are paid less in first-to-default baskets than larger baskets. One justification for these higher correlations is that there is very little demand for selling second-to-default type protection, forcing dealers to hold this risk (or to hold first-to-default protection). But investors rarely feel sorry for dealers, so we don’t give this argument too much weight. Rather, we believe long credit investors are willing to accept higher correlation (lower spreads) on first-to-default baskets because of smaller exposure to “random” defaults than in a 100-name basket. For the more common large basket liquidity points in the market, it may be harder to draw comparisons because portfolio differences can be significant. Yet, we do observe slightly lower implied correlations for first-loss tranches on the larger baskets. A reason for this correlation skew could be a desire by first-loss investors to be compensated for the additional “complexity” of dealing with a larger portfolio.
130
WHAT ABOUT CLOs?
Leveraged loan-backed CLOs tend to have underlying portfolios ranging from 50 to 100 issuers. The correlation models tell us that the absolute sensitivity to portfolio size is lower across the board (compared to synthetic investment grade). We attribute this lower sensitivity to the nature of leveraged loan credit risk (higher default likelihood, higher recovery given default) and the thicker tranches (again compared to synthetic investment grade). The patterns, though, are similar (see Exhibit 4). exhibit 4
CLO Tranches – Similar Pattern but Less Sensitive
Spread as % of 100-Name Basket 110% For senior tranches, risk decreases as the number of credits in the basket increases
108%
0-8%
8-14%
21-28%
28-35%
106%
14-21%
Junior mezzanine tranches are least sensitive
104% 102% 100% 98% 96% For first loss tranches, risk increases as the number of credits in the basket increases
94% 92% 90% 40
50
60 70 80 Number of Credits in Basket
90
100
Source: Morgan Stanley
SOME PRACTICAL IMPLICATIONS
In our model-based analysis above, we took great care in keeping the spread distribution of credits consistent as we moved from small to large baskets. In practice, any two investment opportunities in different sized baskets will also carry differences in credit risk. Even if the average spreads of the portfolios are “comparable,” the distributions of those spreads may not be, and can have a larger valuation impact than the basket size. As such, in many situations, basket sizes can be a second-order effect relative to actual credit differences. For those looking at opportunities involving baskets of different sizes, our advice is to balance the general rules of thumb and the metrics we have described with the actual differences in portfolio credit risk. In particular, credits in the tails of the distribution that do not overlap can be an important driver of relative valuation.
Please see additional important disclosures at the end of this report.
131
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 13
Learning to Live in a Skewed World
April 30, 2004
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Anisha Ambardar Angira Apte
In derivatives markets, the term “skew” is a polite way of saying someone (or something) is wrong. Volatility skew exists because options with different expirations and strikes trade at different implied volatilities. Some blame this on the markets not being efficient, while others claim that the models are not sufficient. In our view, there is truth to both statements, given that skew exists in every derivatives market we can think of. The correlation markets should be no different, since tranches can be thought of as options on a portfolio experiencing losses. Although somewhat smaller than the market experienced late last year, we still see a significant amount of correlation skew in tranches on the benchmark products (see Exhibit 1). There are many drivers of this skew, stemming from both the models and the markets, and even perfect models and reasonably efficient markets will never remove it entirely, because not all investors think alike. In a sense, that’s what makes markets. In this chapter, we attempt to describe at least some of the factors that drive correlation skew today. From a market perspective, there are important technical differences among various investor communities that ultimately drive flows and pricing, and, thus, correlation skew. The models are continuing to evolve, as are ways market participants use them. Both can lead to opinions about model shortcomings, which, in turn, also drives pricing and correlation skew. WHEN DO MARKETS AND MODELS DISAGREE?
There are many technical reasons why the markets can disagree with the models, causing a skew in correlation. Differences in credit analysis methodologies and performance metrics are key reasons. Not all investors live in the same accounting regime, and ratings are much more important to some than for others. For example, correlation skew between the 7-10% and 10-15% tranches can be explained by flows from ratings-based investors (more demand for the former, leading to higher correlation).
132
Living with Correlation Skew – Implied Correlation Across Tranches and Benchmarks
exhibit 1
0-3%
DJ TRAC-X NA 1 20.0
DJ TRAC-X NA 2 21.5
2-5%
DJ CDX NA 2 19.5
33.0
3-7%
15.0
4.0
3.5
7-10%
20.0
17.0
16.5
10-15%
24.0
23.0
22.0
Source: Morgan Stanley
Another big difference between markets and models is that investors do not necessarily have to agree with the risk-neutral approach (i.e., the risk of a credit is fully described by its spread). The lessons we learned in 2002 tell us that investment grade credit can be very asymmetric in nature, even if the ultimate default risk is low. Spreads do not necessarily reflect this phenomenon. Furthermore, risk-neutral models do not incorporate fundamental aspects of credit risk, unlike structural models like Moody’s KMV. 1 Two credits that trade at 50 bp can have very different fundamentals, and can be valued very differently using other types of models. Other factors can force markets and models to disagree as well, including the presence of “story” credits in portfolios, different maturities, or even the number of credits in a basket. HOW ARE THE MODELS INSUFFICIENT? LISTEN TO THE “BUZZ”
A year ago, there was a lot of focus on “mainstreaming” the popular models to introduce a standard “language” in the market. We think these efforts were largely successful and introduced a lot of transparency and liquidity to a market that suffered from a lack of both. Yet, from a modeling perspective, more progress is necessary, and there is certainly a lot of “buzz” today regarding these issues. The two most important, in our view, are differences caused by assuming a single correlation instead of a correlation matrix, and the notion that mezzanine tranches can have more than one implied correlation value for the same price. The first issue is simple to explain but hard to calculate, while the second issue is a bit harder to explain, but much easier to calculate.
1
For a detailed study, see “Valuing Corporate Credit: Quantitative Approaches vs. Fundamental Analysis,” October 2002.
Please see additional important disclosures at the end of this report.
133
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 13
ONE NUMBER VS. A MATRIX – A REVOLUTION?
Simple forms of the standard model assume that the correlation (of default events) for various credits can be summarized by one number, when, in theory at least, an entire matrix is necessary. The practical problem is that it is mathematically very hard to imply a correlation matrix from a market price, especially for large underlying portfolios. However, computational short cuts are possible (sparse and/or partially specified matrices), and the market may begin to embrace this concept at some point. Will the use of correlation matrices in models explain some of the skew we see in today’s single-correlation models? Intuitively, it makes sense, but we do not claim to have all the answers yet. The biggest risk of using average correlation as a substitute for a correlation matrix is when the matrix is disperse (i.e., it has values that are very different from the average). This can be bad for long credit investors in equity tranches, albeit more beneficial to senior tranche investors. As an example, consider a “diverse” basket containing credits from multiple sectors, where credits in one sector are highly correlated to each other, but relatively uncorrelated to credits in other sectors. Relative to a basket in which all credits are equally correlated, the disperse matrix can be bad for equity (but good for seniors) because of the existence of a large region of the matrix with lower correlation values. Using the standard models, equity tranches do trade at lower correlation levels than more senior tranches, so equity’s relative attractiveness may simply be making up for the risk of a disperse correlation matrix. exhibit 2
Correlation Sensitivity Varies with Seniority
bp 900 800
2-5%
3-7%
7-10%
10-15%
15-30% 700 600 500 400 300 200 100 0 10%
20%
30%
40%
50% Correlation
Source: Morgan Stanley
134
60%
70%
80%
90%
TRANCHE EQUIVALENCE: TWO LEGS INSTEAD OF ONE
One of the biggest limitations with the market-standard correlation models is that there is typically one tranche whose spread is insensitive to correlation (the term “correlation invariant” is popular). Such a situation is a correlation trader’s nightmare (not to mention risk managers) because the models are not providing useful information. The problem benchmark tranche in today’s spread and correlation environment is the 3-7% (see Exhibit 2). In different environments it would be other tranches (e.g., 7-10%, if correlation were higher or spreads wider). In the press recently, a market participant introduced a solution to this problem, based on a notion borrowed from the equity markets. 2 Buying and selling options on the same stock or index but with different strikes and/or maturities is popularly termed a “call spread.” When the implied volatilities on the two options are different, the strategy is quoted using both implied volatility measures. Essentially, the investor forms a relative value opinion on each leg of the trade, rather than on the combined position, which may be hard to evaluate because of its complexity. One can think of a mezzanine tranche in the same way. For example, a long credit position in 3-7% is economically equivalent to a short position in 0-3% and a long position in 0-7%. WHY THE COMPLICATION? RELATIVE VALUE INFORMATION
Why is this unusually complicated way of describing a 3-7% tranche useful? It is because of the difficulty in implying a “reasonable” correlation level for this tranche. Breaking the tranche down into two legs (0-3% and 0-7%) that are not invariant to correlation is a way of getting around this problem. Effectively, investors can evaluate two meaningful implied correlation values instead of one meaningless value to form relative value opinions. This process can be “bootstrapped” for other mezzanine tranches, as well.
2
See Derivatives Week, April 12, 2004.
Please see additional important disclosures at the end of this report.
135
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 13
Tranche Equivalence – Long/Short Combinations
exhibit 3
Benchmark Tranche 3-7% 7-10% 10-15%
First Loss Tranche 0-3%
Replicating Portfolio First Loss Imp. Tranche Correlation Position Short 21.2%
First Loss Delta 12.2x
0-7%
Long
27.5%
10.1x
0-7%
Short
27.5%
10.1x
0-10%
Long
31.6%
8.3x
0-10%
Short
31.6%
8.3x
Implied Benchmark Tranche Delta 8.5x 3.9x 1.9x
Source: Morgan Stanley
As an example, a long credit position in 3-7% (Dow Jones TRAC-X) is economically equivalent to 21% correlation on a short credit position in 0-3% and 27% correlation on a long credit 0-7% position (see Exhibit 3). Since both tranches are first loss and, hence, long correlation, the higher correlation on 0-7% tells us that 3-7% is rich relative to the 0-3% tranche (which is at least partially explained above). While we may expect an upward sloping correlation curve, this methodology makes it clear how sloped the curve is for various tranches. WHAT’S THE RIGHT DELTA?
Another important application of this tranche equivalence method is to determine the right delta for a tranche. The delta of the currently correlation-invariant 3-7% tranche is actually very sensitive to correlation (see Exhibit 4). Active market participants concur that a delta of approximately 8 is assumed for this tranche today, which is close to what the tranche equivalence method implies (see Exhibit 5). A correlation level of 4% leads to a delta of 13, implying 36% more protection to be purchased as a hedge – quite a big difference for relative value investors. WHERE DO WE GO FROM HERE?
Two things are clear to us: Models will continue to evolve, because there is still a lot of wood to chop with respect to understanding the behavior of correlated default risk. Correlation skew will persist, nevertheless, because investors think differently about credit risk and operate under different performance measurement regimes. That’s what makes markets.
136
bp
Delta
exhibit 4
3-7% Tranche – The Delta Is Sensitive to Correlation
13
3-7% Delta (Left)
650
12
3-7% Spread (Right)
600
11
550
10
500
9 450 8 400
7 6
350
5
300
4
250 10%
20%
30%
40%
50%
Correlation
Source: Morgan Stanley
exhibit 5
Picking the Right Delta
Implied Correlation 4.0%
Delta @ Implied Correlation 13.3x
Delta Using Tranche Equivalence 8.5x
Delta Difference 4.8x
Delta Difference (%) 36%
7-10%
17.0%
4.6x
3.9x
0.7x
15%
10-15%
23.0%
2.3x
1.9x
0.4x
17%
Tranche 3-7%
Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
137
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 14
Correlation – The Delta Dawn
July 23, 2004
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Anisha Ambardar Angira Apte
It has been well over a year since the market for index tranches started trading, and in that time the correlation space has become one of the fastest growing areas of investor interest and participation within all of fixed income investing. Today’s standardized tranches are an outgrowth of the original CDO market, even though many current investors never participated in this original market. In fact, we find it somewhat alarming that many market participants think of the tranche space as a new market, when in fact the first synthetic CDOs were issued in the late 1990s. If correlation was really a new market, we would be worried, because a good deal of growth has occurred over a short time period and in a relatively benign credit environment. As we have written in past research, we find it very comforting that the synthetic CDO market has been through a credit cycle, although many of today’s “new” investors have not benefited from that experience. The broadening and deepening of the market over the past year, though, is very significant and should certainly not be dismissed by any measure. There has been an explosion in liquidity and an associated boom in volumes, not to mention a whole new generation of investors. Two years ago, there was a limited number of investors using even basic models but today there is a fair amount of investing, trading, and hedging that is tied very closely to the output of correlation models. Further, more than a few investors use even more sophisticated models or approaches. exhibit 1
Spread Duration Decays Over Time, But Not Uniformly
Duration 6.00
0-3% 3-7% 7-10%
5.00
10-15% Index
4.00 3.00 2.00 1.00 0.00 -1.00 5.00
4.50
4.00
3.50
3.00 2.50 Maturity
2.00
1.50
1.00
0.50
Source: Morgan Stanley
We are often asked by investors how the standard tranches have performed relative to both the market and models. In derivatives speak, this is equivalent to quantifying the performance of delta-neutral trading strategies. Some investors have expressed frustration that the deltas were in fact not optimal, causing some performance drag (or gain depending on the strategy), despite the relatively quiet spread environment. We are not
138
surprised by these performance outcomes given the importance of the other significant factors and, fortunately, we now have some historical data to highlight this phenomenon. Our simple performance studies over the past nine months show that delta-neutral positions (selling tranche protection, buying delta-equivalent of index protection) in the four most trafficked tranches resulted in positive performance. Yet, probably more important than the absolute numbers is quantifying the drivers of performance, which we attempt to attribute to the other factors including correlation, spread compression, time decay, and of course carry itself. And since hindsight is the only perfect science, we use some of our historical performance data to demonstrate what the best fit delta would have been. DRIVERS OF PERFORMANCE
In derivatives markets, investors often implement delta-neutral trading strategies to hedge day-to-day market movements, allowing them to focus on other longer-term goals, such as volatility, convexity, or simply positive carry. Yet, deltas do not always explain all of the price movements, especially when moves are big. In the correlation space, other factors can play a significant role, for a variety of reasons. •
Implied correlation itself is probably the most important performance factor, especially when it moves around a lot, particularly for junior tranches (if this sounds obvious, that’s because it is).
•
As we have addressed in detail in previous research, the gamma or convexity of tranches can be significant on a relative basis. At today’s spread levels, 7-10% and 10-15% tranches have a fair bit of negative convexity for wider moves in spread (for more details see Chapter 10).
•
Spread compression (or decompression) can be a driver of performance because it changes the shape of the default distribution (i.e., size of tails). Tranches can be impacted by this in different ways (see Chapters 25 and 27).
•
Even though this is a very model-centric phenomenon, the time decay of various tranches can be vastly different, as we highlight in Exhibit 1. In particular, long credit investors in relatively senior tranches benefit from a fall-off in risk more immediately (for more details, see Chapter 26).
•
Lastly, picking the right delta is more of an art than a science. Deltas for small moves will be very different than deltas for big moves, and we see market participants using both, depending on what type of risk they are trying to hedge. Furthermore, the correlation invariance problem that the industry has focused on is pushing market participants toward using base correlations, which can result in somewhat different but hopefully more accurate deltas, especially for junior mezzanine tranches (see Chapter 13).
DELTA-NEUTRAL POSITIONS RESULT IN PERFORMANCE DIVERSION
In an attempt to illustrate practical investor behavior, we simulated performance of deltaneutral trading strategies in the standardized tranches using the following fairly simple method. We initially created delta-neutral positions (starting in October 2003) in each of the tranches (i.e., sell protection on the tranche and buy the delta-neutral equivalent protection in the index) and then rebalanced these positions on a monthly basis, updating
Please see additional important disclosures at the end of this report.
139
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 14
hedges with the then-appropriate delta (assuming no bid-offer). We used the Dow Jones TRAC-X Series 2 index since it has the longest continuous tranche pricing history. Despite the rather unexciting spread environment (DJ TRAC-X Series 2 widened from 59 bp to 66 bp during this period), delta-neutral strategies had relatively large performance outcomes, where conventional wisdom would have told us otherwise (see Exhibit 2). Interestingly, selling protection in the tranches versus delta hedging in the underlying index resulted in positive performance for almost all of the tranches, ranging from 10 bp for the 10-15% tranche to 420 bp for 0-3%. exhibit 2
Tranche Index 0-3% 3-7% 7-10% 10-15%
Level Oct-03 59 43% 380 120 52
Delta-Neutral Paid Off 1
Jul-04 66 41.30% 365 125 50
Total 0.0% 4.2% 3.3% -0.1% 0.1%
Delta Neutral P/L Carry 0.0% -5.3% -5.5% -2.3% -1.1%
MTM 0.0% 9.4% 8.8% 2.2% 1.2%
1
Percent of tranche notional. Source: Morgan Stanley
ATTRIBUTING PERFORMANCE – THE DELTA DAWN
There are clearly explanations for why delta-neutral trading strategies resulted in performance gains. As we highlighted above, several other factors can determine performance; we conducted a performance attribution study to determine what the most significant factors were over the last nine months. In Exhibit 3 we break down the total return of selling tranche protection (without delta hedging for simplicity purposes) into five components. For example, the 0-3% tranche had a total return of 5.5% during the time period, with positive or negative contributions from the five components. Among the positive contributions were 3.8% from premium (500 bp running for nine months), 3.9% from the correlation move (correlation rose from 18% to 22%, which is beneficial for this tranche), and another 3.0% from the time decay. The negative contributors were spread widening of the index (-4.5%, which is approximately equivalent to the 7 bp of widening multiplied by the duration and the delta) and another 70 bp from the fact that the index decompressed marginally, which is bad for the 0-3% tranche. The 3-7% tranche had directionally similar attributed performance relative to the 0-3% tranche, but the spread decompression was actually a marginally positive contributor, while the correlation move accounted for 190 bp. For the 7-10% tranche, premium was the biggest positive factor. The 10-15% tranche had an interesting attributed return profile, as the less-thought-about factors were the biggest contributors. The tranche earned 60 bp from the time decay (one can see the steep curve in Exhibit 1) and 40 bp from the spread decompression, implying that a lot of the index’s risk moved into the tail that directly impacts the lowest tranches.
140
exhibit 3
Tranche Index 0-3% 3-7% 7-10% 10-15%
Premium 0.8% 3.8% 2.9% 0.9% 0.4%
Tranche Performance Attribution
Tranche Performance Attribution Average Spread Spread Dispersion Correlation Time NA NA -0.5% NA 3.9% 3.0% -4.5% -0.7% 1.9% 1.9% -3.7% 0.5% 0.0% 0.6% -1.1% 0.3% 0.1% 0.6% -1.1% 0.4%
Total 0.2% 5.5% 3.5% 0.7% 0.5%
Source: Morgan Stanley
HINDSIGHT IS THE ONLY PERFECT SCIENCE
If the above attribution analysis presents too many numbers to digest in one sitting, the key take-away is that the deltas on the tranches only explain part of the performance, even during a low-volatility credit environment. Did the models misguide us – were there in fact better deltas that we could have used? To answer this question we ran a simple historical regression of index price movement to tranche price movement, and found that the “best-fit” deltas that came out of the analysis were quite similar to the “average” model generated deltas used in our simulation (except for 3-7%, see Exhibit 4). The R-squared numbers from the regressions were in the 60-70% range, suggesting that they are reasonably strong, but again highlighting that a meaningful portion of performance cannot be captured by index price changes. exhibit 4 Tranche 0-3% 3-7% 7-10% 10-15%
Historical and Model Deltas Say the Same Thing Regressed Delta 11.9 8.7 4.0 1.8
Average Rebalanced Delta 12.0 10.7 4.2 2.0
Source: Morgan Stanley
WHAT ARE WE MISSING?
Significant one-off idiosyncratic risk can be a big driver of performance as well. Spread compression/decompression captures this to some degree, but when there is a real decoupling, then it is all about single-name risk. Although we have not had much of this in the US credit markets over the past year, in previous research we have addressed the impact of the Parmalat default on tranches, which we felt was in line with model predicted behavior (see Chapter 60). DAWN IN A WORLD WHERE THE SUN NEVER SETS
While correlation markets are big, global, and clearly here to stay, transparency continues to improve and we encourage market participants, seasoned or not, to use the opportunity to gain more investing insight. Absolute performance, good or bad, is not always the most meaningful result. How you got there is important as well.
Please see additional important disclosures at the end of this report.
141
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 15
Tranche Steps – One Back, Two Forward
September 24, 2004
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Ajit Kumar, CFA
While the notion that credit portfolio tranches are exposed to correlated default risk is as old as the CDO market itself, investment strategies based on this idea have really only formed over the past year or so, attributed mainly to the liquidity injection from tranche-trading in the benchmark indices. Today the term “correlation trading” is mentioned in many business planning meetings, but the reality is that investment strategies built around the leverage, convexity and liquidity of these instruments are the biggest benefit for most investors. Yet, with the flurry of new instruments across investment grade and high yield markets, this basic message gets a bit lost, in our view. Why are tranche investment strategies interesting? The most important reason is that they allow investors to separate two important aspects of credit risk: default risk and spread risk. In a pure single-name portfolio, both components are closely linked together. With tranches, which are effectively options on the aggregate credit loss severity of a portfolio, it is possible to separate idiosyncratic default risk from marketwide spread movements. In the current credit environment, we are supportive of taking default risk but are not very excited about the actual level of credit spreads. Deltaneutral long/short trades involving subordinate and senior tranches can implement this view, but the details are essential. For several reasons, we feel that it is important to take a step back and look over the usefulness of the correlation market today, from a strategic perspective. First, the market continues to grow daily with many new participants ready to manage credit risk in this new fashion. Second, with a fairly complicated menu of instruments trading, investment strategies abound, and these differ between investment grade and high yield. Third, the duration of an investment strategy is as important as the strategy itself, and today there are quite a few choices. Finally, from a relative value perspective, the concept of correlation has now become almost a purely technical term, with market participants focused on base correlation, which has a very different intuitive feel than real default correlation. OPTIONS ON DEFAULT – 101
The basic idea behind tranching credit portfolios is to control the amount of idiosyncratic default exposure. Tranches that “feel” like they will have losses due to default over the maturity are effectively in-the-money options on default. Similarly, tranches where losses due to default over the term seem unlikely resemble out-of-themoney options. In market-standard correlation models, the factor that determines whether a tranche is in- or out-of-the-money is the level of spreads (along with spread distribution). From a bottom-up perspective, an investor’s actual opinion on defaults for a given portfolio is an important source of relative value, as it may differ from the risk-neutral model’s view.
142
TRANCHE CHARACTERISTICS
If one thinks of tranches as options on default, then understanding their sensitivity to spread movements (or defaults) becomes much easier. Images of convexity and hockey sticks should come to mind, and we highlight four of these characteristics in Exhibit 1 for today’s benchmark instruments. In particular, we focus on tranche sensitivity to small market-wide spread moves (DV01 or delta), large market-wide spread moves (convexity or gamma), time decay (roll down or theta), and individual defaults (idiosyncratic risk). The numbers are tranche sensitivity measures relative to the underlying index. exhibit 1
Index 5Y IG
Tranche 0-3%
Tranche Sensitivity Menu
Delta 14.0x
Time 1 Decay 0.16%
Downside 2 Convexity 0.5x
3
Avg. Default PV 34.7x
5Y IG
3-7%
7.6x
0.17%
1.0x
5.6x
5Y IG
7-10%
3.3x
0.19%
1.3x
0.2x
5Y IG
10-15%
1.4x
0.22%
1.6x
0.4x
5Y IG
15-30%
0.5x
0.21%
1.8x
0.0x
10Y IG
0-3%
5.4x
-0.28%
0.4x
29.0x
10Y IG
3-7%
8.3x
0.06%
0.6x
11.3x
10Y IG
7-10%
5.3x
0.22%
0.9x
4.2x
10Y IG
10-15%
3.2x
0.25%
1.1x
1.1x
10Y IG
15-30%
1.3x
0.30%
1.5x
0.1x
HY
0-10%
2.0x
2.96%
0.6x
8.4x
HY
10-15%
2.9x
2.92%
0.8x
4.3x
HY
15-25%
2.7x
0.76%
1.0x
3.2x
HY
25-35%
1.4x
1.21%
1.2x
2.2x
HY
35-100%
0.0x
1.95%
1.7x
0.0x
1
1 year, delta adjusted DV100 / (100*DV01), index multiple 3 Index multiple Source: Morgan Stanley 2
For example, from the perspective of a seller of protection, a 7-10% tranche on the 5 year investment grade CDX index has a DV01 or delta of 3.3x relative to the index and is negatively convex in that a 100 bp move in spread results in a price impact that is 1.3 times more convex than that for the index itself. The 7-10% does benefit from roll down/time-decay more so than a delta-neutral amount of the index (0.19% in price terms versus 0% for the index) and has less exposure than the index to the first default (by a factor of 0.2x, assuming the average default impact of the widest and tightest credits). STRATEGIES FOR TODAY – SLICING AND DICING
In previous research, we have described in detail some of our basic strategies in the correlation space (see Chapters 24 and 26). In this chapter, we summarize our core view on the market as being supportive of taking default risk but not very excited about the level of spreads. We find a handful of tranche strategies attractive today.
Please see additional important disclosures at the end of this report.
143
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 15
Long default risk, neutral to short spreads, positive carry. This strategy can be implemented by selling protection on any tranche with a large exposure to individual defaults (particularly discount priced first-loss tranches), combined with delta-neutral or net DV01 short hedges in the underlying index/portfolio or junior and senior mezzanine tranches. Protection in the mezzanine tranches is advantageous because they have more spread exposure than individual default exposure and will benefit from convexity when spreads widen. Some of the best tranches to use today include fiveyear investment grade 7-10% or 10-15% or 15-25% of HY CDX. The positive carry on the trade comes from the first-loss tranche’s larger idiosyncratic risk relative to protection in the index or more senior tranches. We also recommend selective singlename hedging to either implement negative credit views or mitigate jump-to-default risks. Long spread convexity. The simplest way to isolate spread convexity is to buy protection in tranches where convexity is relatively high (7-10% and 10-15% in fiveyear IG indices), and delta hedge with the underlying index. In today’s markets, many of these strategies will be positive carry as well, but they may suffer from a time-decay or roll down phenomenon. Also, the long convexity trade does have increased idiosyncratic exposure through the delta hedge. Playing economic cycles. Credit curves are now fairly steep out to 10 years, relative to when we first addressed the idea of mixing short and long maturities in correlation strategies. Nevertheless, selling protection in first-loss tranches versus buying protection in mezzanine tranches can be implemented with different maturities to express a cyclical view. In general, we favor selling first-loss tranches to shorter dates and buying more senior protection to longer maturities. Strategies that are DV01 neutral today can actually become negative DV01 over time, given the time decay phenomenon. CORRELATION CAN BE CONFUSING
Since the liquidity infusion in the correlation market over a year ago, perhaps one of the biggest shifts in thinking has been about correlation itself. For pricing and relative value purposes, the Street has moved toward using base correlation instead of compound correlation because the standard models are better-behaved this way (for a full description, see Chapter 13). As we have mentioned in the past, we continue to view correlation from a fairly technical perspective, although the original sensitivity concepts do not change. Base correlation for a given tranche is the implied correlation for a hypothetical tranche that attaches at 0% (the base) and detaches at the detachment point for the given tranche. For example, the base correlation for the 3-7% tranche is the implied correlation for a 0-7% tranche. However, since 0-7% does not really trade in the market, the price for the 0-7% tranche is computed by “adding” together the prices for 0-3% and 3-7%. In the base correlation framework, all tranches are equity style, meaning that they appreciate in price as correlation rises (i.e., long correlation). This makes relative value a bit simpler, and most focus on the skew (or curve) in correlation between the tranches.
144
Before we delve into relative value, we would like to make one point very clear. Base correlation is clean and convenient, but it does confuse the true notion of default correlation. Tranches’ sensitivity to realized default correlation does not change, i.e., the junior tranches tend to be long correlation (diversity is bad for them) and the senior tranches tend to be short correlation (diversity is good for them); see Chapter 9 for our early thoughts. From a day-to-day relative value perspective, base correlation is the cleanest metric, but in terms of intuition on the risks inherent in tranches, compound correlation remains a valuable tool. CORRELATION RELATIVE VALUE
To help in relative value analysis, we have applied the base correlation calculations to about one year of benchmark tranche pricing history. One of the easiest ways to use this data is to focus on the shapes of the base correlation curves. The curve for the 37% tranche has steepened from near 0% at the end of the year to about 10 correlation points today (0-3% trades at 20% correlation and 3-7% trades at 30%, see Exhibit 2). We consider this a significant move and have discussed the relative richness of junior mezzanine notes recently (see Chapter 43). The 10-year 3-7% tranche followed a similar pattern, but does not appear to be as rich, from a base correlation perspective. exhibit 2
Richening Mezzanine Notes – Base Correlation Curve for 37% Tranche
bp 15%
5Y 3%-7%
10Y 3%-7%
HY10% - 15%
10%
5%
0%
-5% Oct-03
Dec-03
Feb-04
Apr-04
Jun-04
Aug-04
Source: Morgan Stanley
CONCLUSION
While we have spent a considerable amount of our research efforts over the past two years discussing the intuition behind and sensitivity of correlation instruments, we continue to view this space as novel from a credit portfolio management perspective. Furthermore, our market sense is that quite a few investors are on the verge of getting involved over the near term, as investments in analytical infrastructure come to fruition. Therefore, revisiting some of the basic steps should help to outline long-only and long/short investment strategies.
Please see additional important disclosures at the end of this report.
145
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 16
High Yield – Thinking Outside the Correlation Box
November 19, 2004
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Ajit Kumar, CFA
A key reason why we think the high yield structured credit opportunity is an interesting one revolves around the prospect of juxtaposing fundamental and quantitative approaches to valuing high yield credit risk. More so than in the investment grade markets, it is important for high yield structured credit players to think outside of the correlation box, and then to use market pricing and correlation intuition as a way to calibrate their views. We first argued for this approach in a previous report, where we described some of our fundamental views on the 100 names that make up the HY CDX portfolio (see Chapter 29). In a market where underlying cash bonds frequently trade on a dollar price basis, and investors often think about default and recovery risk separately, the opportunity to compare fundamental views with the results of riskneutral models is compelling. One way to better understand this phenomenon is to observe the apparent pricing “discrepancies” between investment grade and high yield equity tranches. While the tranches are far from being equivalent from a price or expected loss perspective, they trade at very different implied correlation levels in the market (19% for 0-3% investment grade tranche versus a much higher 39% for 0-10% high yield tranche). Furthermore, the correlation skew among all of the standard investment grade tranches is much steeper than in high yield. From a protection seller’s perspective, the lower base correlation on an equity tranche implies that the risk is cheaper while a lower skew implies the same for more senior tranches. Yet, in the still burgeoning market for synthetic high yield, it is hard to imagine that equity tranches can really be that rich relative to investment grade so quickly. Is there a better explanation? We do not claim to have all of the answers, but we do offer up several simple and fundamental explanations for why correlation levels are indeed much higher in high yield vs. investment grade tranches, in our view.
146
1.
Statistics favor high yield. At 325 bp on HY CDX, the market implies a fair number of defaults (approximately 15% losses) over five years. There are clearly many bullish scenarios where the market (or a specific portfolio) could experience much fewer defaults, especially given high yield fundamentals. The existence of these scenarios favors high yield equity tranche investors, relative to investment grade. Furthermore, given liquidity, high yield equity tranche investors could feel more comfortable about getting out of their risk, especially compared to the much less liquid cash CDO world.
2.
Recovery outcomes may favor high yield, ironically. In a market where investors differentiate default risk and recovery values, there is an argument that recovery values for HY CDX names could be higher than typical assumptions or market averages, given sector biases. Also, since it takes more defaults to cause pain in equity tranches, the likelihood is small that all of these defaults experience low recoveries. Furthermore, historical data show that when default rates are low, recoveries tend to be high, and vice versa. This makes equity tranches much more extreme plays on default risk than correlation models suggest.
3.
Defaults are more correlated in high yield. High yield defaults have historically been more concentrated (in sectors) than investment grade, which is an argument for higher default correlation for equity tranches.
4.
Standardized terms favor high yield equity tranches. Finally, given mechanics of equity tranches, the best and worst case scenarios for HY equity tranches are more positively skewed than for investment grade equity tranches.
There is clearly a short story, if not a whole book, behind each of these themes, but we will attempt to shed a bit of light on them in a much more abbreviated form, with the goal of gaining a bit more intuition into high yield correlation. 1. STATISTICS FAVOR HIGH YIELD EQUITY TRANCHES
Our first point above is really about statistics. Exhibit 1 illustrates the inherent volatility in 5 year cumulative default experience for both the high yield (using a HY CDX-like ratings distribution, which is 50% BBs) and investment grade markets over time. The key takeaway is that high yield cumulative defaults are much more volatile over time than investment grade, both in an absolute sense and relative to the attachment points of the typical index tranches. Depending on the time period, a portfolio like HY CDX (in terms of ratings distribution) could generate losses only in the first tranche (0-10%) or could have sufficient losses to penetrate the bottom three tranches. For investment grade, losses generally are contained within the first tranche (0-3%). Additionally, the sampling risks inherent in a 100 or 125 name portfolio (relative to the broader market) should add to this volatility making more extreme results for portfolios possible.
Please see additional important disclosures at the end of this report.
147
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 16
exhibit 1
%
HY Default Rates Are Volatile – HY CDX-Like Ratings Distribution
25 IG Losses HY Portfolio Losses
20
15
10
5
1999
1998
1997
1996
1995
1994
1993
1992
1991
1990
1989
1988
1987
1986
1985
1984
1983
1982
1981
1980
1979
1978
1977
1976
1975
1974
1973
1972
1971
1970
0
Source: Morgan Stanley, Moody’s
At 325 bp on HY CDX, derivative models imply roughly 15% losses over five years, which is at the high end of historical five-year loss rate for a portfolio with similar ratings distribution to the index. Fundamentally, the default risk in the index feels better than this today (given cash on balance sheets, see Chapter 29). If we are headed for a trough in the HY default risk cycle, the statistical argument above favors the high yield equity tranche investor, since it has the most room to run in terms of absolute improvement in default experience. Even if fundamentals worsen, the liquidity in the tranche market available to high yield equity tranche investors may be a relative advantage to playing the asset class directly and is an option that never existed in the cash CDO world. 2. RECOVERIES MAY BENEFIT HIGH YIELD EQUITY TRANCHES
On a very basic level, equity tranche investors in investment grade are exposed to more “random” event risk than high yield equity investors. This randomness comes both in the form of default events and uncertainty of recovery. Consider the maximum potential impact of varying recovery from 0% to 40% for a single default on the equity tranches. For the 0-3% tranche of a 125 name portfolio the impact of a single default ranges from 20% to 27% of notional while for the 0-10% tranche of a 100 name high yield portfolio this same percentage ranges from 6% to 10%. Additionally, the increased default risk in high yield relative to investment grade actually serves to further dampen the effect of any single recovery outcome on a high yield portfolio as it will be a smaller portion of the total expected losses. For investment grade, a single default event contributes more to the aggregate losses of the portfolio and hence individual recovery differences matter more.
148
Another relevant point about recoveries is that, based on historical data, low default rate environments tend to result in high recoveries on average, while high default rate environments tend to result in low recoveries. The economic argument for this is that asset values go down when lots of asset “supply” hits the market. This is particularly relevant in high yield where high default correlation in a particular sector (which we discuss below) could also drive lower recovery in that same sector. To the extent default experience is driven by company specific events rather than sector economics this is more beneficial for high yield. 3. DEFAULTS MAY BE MORE CORRELATED IN HIGH YIELD
After discussing the notion of default correlation in a technical sense for so long, we finally get an opportunity to address it in a truly intuitive manner, thanks to the high yield opportunity. In investment grade, it is hard to argue that a given portfolio has high or low default correlation, because the credits themselves are so far from default, and historical experience suggests that investment grade defaults are not well correlated to sectors, which are big drivers of index/portfolio design. In high yield, we have seen many sectors over time experience much higher defaults than the larger market, including Telecom during the last credit cycle (see Exhibit 2). In addition to the increased sector correlations in high yield, a structural argument can be made for higher correlation. If we consider that, generally speaking, high yield companies are closer to the default barrier than their investment grade brethren. And, if we are willing to accept the fact that macroeconomic factors affect asset values in both markets then the relative effect should be greater in high yield. This argues for higher absolute correlations in high yield.
Please see additional important disclosures at the end of this report.
149
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 16
exhibit 2
Defaults Have Sector Themes – Default Correlation Is Higher
Hotels, Casinos, & Gaming
2001 2002
Nondurable Consumer Products
2003
Leisure, A musement, & Textiles, Leather, & A pparel Healthcare, Education, & A utomobile Media, Publishing, & Broadcasting Forest Products & Paper Retail Energy Beverage, Food, & Tobacco Chemicals, Plastics, & Rubber Electronics Metals & Mining Finance, Insurance, Real Estate Telecommunications 0
Source: Morgan Stanley, Moody’s
150
5
10 15 20 25 % of De faulting Is s ue s
30
35
4. THE ASYMMETRY OF STANDARDIZED TERMS
The standard terms for equity tranches makes it very easy to determine the protection seller’s best and worst case scenarios (see Exhibit 3, which is based on cash flows after protection is sold). For example, in the 0-10% HY tranche, where investors receive 75 points upfront for selling protection, the maximum downside is 25% of notional, while the maximum upside (zero defaults) would be 75% (seller of protection keeps the entire up front payment). For the 0-3% IG tranche, it is a similar example (investors receives 34 points up front), but since there is 500 bp of running premium the maximum upside is roughly 51% of notional while the downside is 66%. exhibit 3
IG 5Y 0-3%
Max Upside/Downside Comparisons of Equity Tranches
Max Upside 51%
Max Downside 66%
Upside/Downside Ratio 0.8x
IG 10Y 0-3%
74%
46%
1.6x
HY 5Y 0-10%
75%
25%
3.0x
HY 5Y 10-15
39%
61%
0.6x
Source: Morgan Stanley
Both of these examples stand in stark contrast to what is typical for investment grade single-name investors who can face large (though unlikely) downside scenarios and small upside scenarios. Most notable on this front is the 0-10% tranche of HY which, because of large upfront payment, has 3 times the upside in a no default scenario. The leverage inherent in this structure is a product of the standardized terms of the traded instruments and presents an interesting upside/downside opportunity for some investors, even though the tranche is very much an in-the-money option on default based on market-implied (or “expected”) losses. Another effect is the implied subordination in the equity tranches generated by the size of the upfront payments. The seller of protection in the 0-10% tranche is only exposed to losses in excess of 7.5% (12 defaults at 35% recovery) of the index over 5 years (because he/she gets paid approximately 75 points upfront, and 0 bp running). In the case of investment grade the seller of equity protection is exposed to portfolio losses (based on the upfront payment) after just one or two defaults. Interestingly this is approximately the ratio of default rates implied by current investment grade and high yield spread levels.
Please see additional important disclosures at the end of this report.
151
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 17
Structural Evolution – Survival of the Fittest?
March 18, 2005
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Ajit Kumar, CFA
Over the past six months or so, there has been an important thematic shift in the structured credit market, in our view. No one will dispute the argument that the standardization and liquidity attributes of index tranches were largely responsible for bringing structured credit into the mainstream. We would argue that the tranche index market is now well established and serves as both the most visible part of the market as well a relatively easy entry point for many investors. In fact, with idiosyncratic events recently in General Motors, Liberty Media, IPG, and Viacom, it is easy to see the impact on leveraged tranches (see Exhibit 1). exhibit 1
DJ CDX 5 Yr
Current Level (bp*) 46.0
Tranche Transparency Is Very Helpful – Impact of Recent Idiosyncratic Events
1 Week Chg (bp*) 3.5
Delta-Implied Chg (bp*)
Change in Correlation Skew (%)
Richer or Cheaper? Cheaper
0-3%
31.5
3.9
2.7
-0.65
3-7%
172.0
17.5
23.5
0.35
Richer
7-10%
58.0
5.5
8.8
-0.2
Cheaper
10-15%
20.4
4.1
3.3
-0.2
Cheaper
15-30%
8.3
1.1
1.4
-0.8
Cheaper
Note: For 0-3% tranche price and price changes are stated in points upfront. Source: Morgan Stanley
As we addressed in an earlier report, the structured credit market was a largely a bespoke one that certainly needed boosts in liquidity and transparency to make it more viable and mainstream (see Chapter 46). This benchmark liquidity was not only important for visibility purposes, but it also has helped the bespoke market in important ways. The resulting flood of structural innovation today is both interesting and rather complicated. The key question is how useful (or not) the innovation really is, and the answers largely depend on who (in the investment community) is asking. In many ways, the deluge of innovation reflects the diversity of credit investors in the market and their demand for sophistication, which we would argue is the important thematic shift in the market.
152
In Exhibit 2, we summarize some of the key innovative features that have either become well established recently, or have been just introduced and show (in our view) some promise. We would argue that in many cases, new features are actually a reaction to some type of risk that was exposed previously, such as excessive idiosyncratic risk during the last credit cycle or associated low recoveries. But others are strictly innovative, including things like interest rate hybrids and principal protection techniques borrowed from other markets. One of the innovations that caught our attention is the use of cross-subordination in CDO-squared structures, and we attempt to provide a bit of insight in this chapter into how valuable it can be. But for those who are not familiar with this new structural feature, a bit of a history lesson on the evolution of CDO-squared structures is necessary, which we happily provide, since we like telling stories anyway. exhibit 2
Key Innovations, or Those That Show Promise
Key Innovation Hybrid CDO Squared (w/ABS)
Our Thoughts Reaction to the over-levered CDO-squareds of several years ago. The large ABS portion serves to delever the structure
Pure CDO-Squared w/Thicker Tranches
Similar idea to yesterday’s CDO-squareds, but thicker tranches serve to delever the structure
Large Collateral Pools w/Less Overlap
Investors have recognized the importance of overlap in CDOsquareds, particularly for subordinate tranches in the outer CDO. Lower levels of overlap are now common
Fixed Recoveries
A reaction to the low recoveries from fallen angel defaults in 20012002. Fixed recoveries reduce uncertainty, particularly in mezzanine tranches, but they are a double-edged sword as IG recoveries can be high as well
Dynamic Principal Protection (CPPI) Application of an age-old technique to credit for investors seeking principal protection. Long/short tranche strategies or hedge fund investments seems to be the natural application Cross-Subordination
An interesting concept that allows subordination within inner tranches of CDO-squareds to be shared
Interest Rate Hybrids
A natural progression as funded credit investors demand differing exposure to interest rate risk, beyond strict fixed or floating rate instruments
Adjustable Subordination (or Adjustable Coupon)
A viable technique for meeting investor demand for managed structures while allowing deal arrangers to hedge risks
Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
153
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 17
CDO-SQUAREDS – A BRIEF HISTORY LESSON
Synthetic CDO-squared transactions have become popular in the market for a variety of reasons, and can play some interesting strategic roles in credit portfolios (for some of our earlier thoughts, see Chapters 11 and 63 and “The Far Side of a CDO-Squared,” June 10, 2004). From the perspective of innovation in the marketplace, they deserve special attention because of their relative size and the story behind them. Early CDO-squared transactions were quite different from today’s state-of-the art technology, demonstrating the learning process that market participants went through. The first deals were issued in 1999 and were generally managed cash flow structures comprised of a large number (80 or so) of already managed junior and senior mezzanine tranches of high yield CBOs, and to some degree CLOs. In retrospect, they were quite levered transactions with a fair amount of concentrated (overlapping) single-name exposure. Most performed poorly when the credit cycle turned and taught investors some important lessons. Today’s CDO-squared structures are different along several dimensions. First, they are generally static synthetic structures. Second, the inner CDOs generally do not comprise 100% of the portfolio; there is often a significant portion of ‘funding’ in the structure, usually through the form of AAA-rated ABS collateral, which serves to delever the structure. As a result, ‘first loss’ tranches in the outer CDO generally carry investment grade ratings. Third, the inner CDOs tend to be senior mezzanine or junior senior tranches of investment grade portfolios. Finally, the transactions are relatively transparent, meaning that investors have complete information on credit exposure and credit overlap (which may not have been the case in the past). It should be noted that CDO-squared transactions where the inner CDOs make up 100% of the collateral do still exist, but typically they include relatively thick tranches (e.g., 10% for investment grade portfolios). Also, in the cash CDO markets, we recently noticed the pricing of CDO comprised of leveraged loan CLO tranches crossing the tape. WHAT’S THE MOTIVATION? SYSTEMIC VS. IDIOSYNCRATIC RISK
Most investors who are attracted to CDO-squared structures are not looking to take on idiosyncratic risk, as there are better ways of doing that. Yet early transactions had plenty of idiosyncratic risk because inner CDO attachment points were low (on HY portfolios) and overlap was relatively high. Today’s structures reflect a flight from idiosyncratic exposure, but leverage is still an important part of the trade. While even the first-loss tranches of CDO-squareds are far from default, they are levered plays on systemic risk. Investors are effectively selling credit convexity, much like those who sell deep out-of-the-money options.
154
STILL MORE EVOLUTION – CROSS-SUBORDINATION
Today’s highly engineered CDO-squared transactions are still evolving, as investor demand continues to get more sophisticated. The overlapping credit issue is transparent now, and we would argue that investors spend quite a bit of time focused on it. The other issue that has come up is the opposite of too much overlap, or too little overlap. In structures where there is not much overlap, the inner CDO tranches are somewhat independent of each other. As such, a natural scenario to consider is one where defaults are concentrated in a single inner tranche, to a point where it then triggers a loss in the outer CDO. The subordination in the other inner CDOs is effectively wasted in this scenario, so investors have wondered if this subordination can be moved over to the inner tranche that needs it. The technique is called cross subordination and implies that subordination of inner tranches can be shared and therefore must be fully exhausted before the outer CDO experiences a loss. On the surface, cross-subordination sounds intuitively appealing, almost like there is nothing to lose by having it. But the real question centers on how valuable it is. The answer is that it depends on how much overlap there is in the portfolio, and the correlation regime we live in. CROSS-SUBORDINATION – WHAT IS IT WORTH?
To get a sense for the value of cross-subordination we conducted a simple simulation (similar to market-standard correlation models) based on a hypothetical CDO-squared structure for which the inner CDO tranches were comprised of six 4-6% tranches of portfolios with spread distributions similar to the current Dow Jones CDX portfolio (average premium of 47 bp). These inner CDOs make up 12% of the notional amount for the outer CDO structure, with high-quality ABS assumed to make up the balance of the portfolio. We examine the impact of adding cross-subordination to the structure by measuring the probability of a principal loss on the first loss tranche of the CDO squared, assuming two default correlation scenarios (5% and 20%). We compare the probabilities of a principal loss for structures with and without crosssubordination to arrive at an implied reduction in default probability derived from the cross-subordination. Exhibit 3 illustrates the results of the study for several different levels of overlap in the underlying six portfolios, where the overlap percentage is defined as the percentage pair-wise overlap between any two of the inner CDO portfolios.
Please see additional important disclosures at the end of this report.
155
Structured Credit Insights – Instruments, Valuation and Strategies
90%
exhibit 3
Cross Subordination Remains Valuable as Overlap Rises (20% Correlation)
Reduction in Probability of Principal Loss
chapter 17
80%
0% Outer CDO Attachment
70%
2% Outer CDO Attachment 4% Outer CDO Attachment
60% 50% 40% 30% 20% 10% 0% No Overlap
12% Overlap
24% Overlap
36% Overlap
Source: Morgan Stanley
The results suggest that for first-loss CDO-squared tranches, there is significant value in cross-subordination structures with probabilities of principal loss reduced by 42% to 56% for the first loss tranche, depending on the amount of overlap (assuming 20% correlation, which we believe is a reasonable assumption). While overlap in the inner CDO portfolios slightly reduces the impact of the cross-subordination, the benefit is quite robust even for fairly extreme overlap scenarios. The value of the crosssubordination decreases as we go higher in the capital structure of the CDO squared. In Exhibit 4 we extend the analysis for an underlying portfolio with default correlation of 5% and we find cross-subordination to be even more valuable, with the probability of principal loss reduced by 76% to 81% for the 0% attachment tranche, depending on the level of overlap. 90%
Cross Subordination Is Even More Valuable in Low Correlation Scenarios (5%)
Reduction in Probability of Principal Loss
exhibit 4
80% 70% 60% 50% 40% 30% 20% 10%
0% Outer CDO Attachment 2% Outer CDO Attachment 4% Outer CDO Attachment
0% No Overlap
Source: Morgan Stanley
156
12% Overlap
24% Overlap
36% Overlap
One point worth noting in the two correlation scenarios above is that the impact of overlap differs as we go up the capital structure of the CDO squared. In Exhibit 5, we show the difference in cross-subordination benefit between the no overlap scenario and the 36% overlap scenario under both correlation assumptions. For the 5% correlation scenario, the difference in the benefit of cross-subordination between the 0% and 36% overlap scenarios increases with subordination while for the 20% correlation scenario the difference in the benefit of cross-subordination decreases with subordination. The implication is that the impact of overlap on the value of the cross subordination is affected by the correlation regime we experience or, in plain English, the severity of the next credit cycle. SURVIVAL OF THE FITTEST
Financial engineering in today’s credit markets reflects Darwinian theories, in our view. The market will ultimately be able to distinguish good ideas from bad, and only the former will survive, but it will take some work to get there. We are strong supporters of continuing structural evolution, especially when it results from previous experience, but we advise investors to maintain a discerning approach. exhibit 5
Cross Subordination Sensitivity to Overlap Differs by Correlation Regime
Benefit Difference Between 0% and 36% Overlap 18% 16% 14% 12% 10% 8% 6% 4%
20% Correlation
2%
5% Correlation
0% 0% Attachment
2% Attachment
4% Attachment
Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
157
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 18
Is Seven Closer to Five or Ten?
July 22, 2005
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj
Much like the development of a term structure in the underlying single-name credit derivatives markets over the past two years, there has been a reasonable amount of activity along the curve in the on-the-run corners of the structured credit space, as well. A diversity of maturities is nothing new in the structured credit world, as more than a handful of early synthetic CDOs (2000-2002) were 7-year structures, and a good amount of the synthetic CDO activity in 2004 was also longer-dated in nature. Yet, there has always been preference for 5-year maturities in synthetic structured credit, as equity and mezzanine-tranche investors do worry (justifiably so) about the incremental default risk in longer maturities. Much of the recent rapid growth in the index tranches has been centered around the 5-year point, as well, and the spring auto-induced volatility in the index tranches certainly demonstrated how much structured credit risk in the market is concentrated in 5-year maturities. As we addressed in our report, “Checking the Pulse of the Market, Post-Volatility,” July 15, 2005, we feel that the environment for longer-dated structured credit bid is ripe. Our reasons include a positive basis in this part of the curve, continued steep credit curves, historically tight spread levels on 5-year mezzanine tranches, and newfound activity in on-the-run 7-year index tranches. Furthermore, the recent increased activity in managed synthetic CDO structures may also be supportive of 7year structures, as the longer term gives managers a better opportunity to manage underlying portfolios, given the increased liquidity expectations as the underlying default swaps approach the 5-year maturity two years down the road. We take a first look at the 7-year opportunity in tranches in this chapter, using the extremely valuable pricing information we obtain from the index tranche markets. While the more mature liquidity points in the 5- and 10-year index tranches provide important relative value benchmarks, whether the 7-year opportunity resembles one or the other depends on multiple factors, as we discuss in detail in this chapter. Our key points are as follows:
158
•
The 7-year 3-7% tranche could have an identity crisis. In a market where there is a hunt for yield, it may benefit from tighter spreads, but it is a cuspy tranche with unstable ratings when the unexpected occurs.
•
Despite tight spreads, the 5-year 7-10% tranche continues to look attractive on a risk-premium basis relative to both historical default rates and model-generated deltas.
•
When considering risk premium adjusted for deltas and historical default risk, mezzanine tranches, including the 5-year 3-7%, the 7-year 3-7% and 7-10%, and the 10-year 7-10%, are the least attractive, while equity tranches and more senior tranches fare better.
•
The 7-year 0-3% currently has the most amount of risk premium of any of the standard tranches relative to deltas and historical default rates, marginally beating out the 5-year 0-3%.
WHAT IS THE MARKET SAYING?
We start by examining the tranched market through the lens of a correlation model. In Exhibit 1, we show recent market pricing for the CDX IG 4 series of tranched products, as well as the present value of losses as implied by the market pricing. A few of these data points offer interesting insight into how the market is pricing risk. exhibit 1
Pricing and Market Implied Losses
Index Tranches Pricing (bp) CDX IG 4
5-Year 56
7-Year 66
10-Year 79
0-3%
44.75%
57.63%
63.75%
3-7%
122.5
279.5
647
41
58.5
140.5
20.5
34.5
74.5
13.75 23.5 PV of Market Implied Loss (% of Notional) 2.4% 3.8%
37.5
7-10% 10-15% 15-30% CDX IG 4
6.0%
0-3%
59.0%
73.8%
81.1%
3-7%
5.3%
15.7%
44.2%
7-10%
1.8%
3.4%
10.8%
10-15%
0.9%
2.0%
5.8%
15-30%
0.6%
1.4%
3.0%
Source: Morgan Stanley
For the index itself, the current curve shape implies 2.4%, 3.8% and 6.0% losses for the 5-, 7- and 10-year maturities, respectively. These implied index loss levels can provide insight into tranche pricing. For example, current index spreads make 5-year 3-7% tranches appear out-of-the-money (expected losses less than the 3% attachment point); these tranches have thus recently behaved more like mezzanine or senior tranches. Ten-year 3-7% tranches appear in-the-money (expected losses of 6% easily penetrate the 3% attachment point) and, for the most part, have behaved more like equity tranches recently. Meanwhile, the 3-7% 7-year tranche is on the cusp of the equity and mezzanine divide, with expected losses almost penetrating a 4% attachment point. While we have limited price history to observe, we surmise that these tranches will behave more like 5-year tranches provided spread levels remain near current levels; however, the risk in these tranches is that they will behave more like their 10-year counterpart in wider-spread market environments. For all but the equity tranches, there is significantly more risk priced into years 8-10 than in years 1-7. For example, both the 3-7% and 7-10% tranches price twice the default risk in years 8-10 than in the first 7 years. This is generally consistent with the subordination that these tranches benefit from, but it also raises the question of how this compares to the patterns of investment grade defaults over time.
Please see additional important disclosures at the end of this report.
159
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 18
RATINGS, CORRELATION AND MARKET TECHNICALS
Another insightful exercise is to observe skew in base correlation values among the various tranches, which can give us a good indication of technicals related to, among other things, expected tranche credit ratings (see Exhibit 2). The base correlation skew for the 3-7% tranche declines as maturity rises, meaning that it gets cheaper relative to the 0-3% tranche. We believe that this is related to the technicals associated with ratings: a 5-year 3-7% tranche is easily BBB-rated, while the 7-year likely straddles the investment grade/high yield divide and the 10-year would be strictly below investment grade. The skew of the 7-10% tranche marginally increases with maturity as it maintains relatively high ratings, and becomes more attractive as tranches beneath it become riskier, again from a ratings perspective. exhibit 2
3-7%
Base Correlation Skew and Ratings Technicals 5-Year 21.15%
7-Year 18.25%
10-Year 9.40%
7-10%
10.35%
12.30%
12.35%
10-15%
12.95%
14.50%
14.55%
15-30%
23.45%
25.05%
27.45%
Source: Morgan Stanley
As an aside, while the market indeed benefits from standardization along many dimensions, standardized tranches force us to think about default risk almost on an arbitrary basis. The original standardized 5-year investment grade index tranches were created by taking into consideration expected ratings on 5-year synthetic CDOs (in early 2003). The fact that the same standard tranches are used for other term structures is almost meaningless, given significantly different ratings and sensitivity to ratings. Again, the 7-year 3-7% is a prime example, straddling investment grade and high yield ratings. WHAT IS THE REAL WORLD TELLING US?
Using a rich set of default data from Moody’s, we show in Exhibit 3 the implied loss distributions for a CDX-like portfolio (in terms of ratings and size) for 5-, 7- and 10-year maturities (see Chapter 32 for a description of how these distributions are generated). A key takeaway from this analysis is the relatively concentrated nature of expected losses for both 5- and 7-year periods, relative to 10-years. In fact, the more evenly distributed losses over the 10-year period demonstrates the increased risk (or uncertainty) of higher attachment points in 10-year tranches, which is a justification for flatter base correlation curves. At least from this perspective, the 7-year loss distributions look and feel more like that of the 5-year maturity; however, the standard structure of the tranches can complicate this analysis.
160
exhibit 3
50%
Historical Loss Distributions – 5- and 7-Year More Clustered, 10-Year More Uncertain
5Y
45%
7Y
40%
10Y
35% 30% 25% 20% 15% 10% 5%
0. 0% 0. 0.5 5% % 1. 1.0 0% % 1. 1.5 5% % 2. 2.0 0% % 2. 2.5 5% % 3. 3.0 0% % 3. 3.5 5% % 4. 4.0 0% % 4. 4.5 5% % -5 5. 0% .0% 5. 5.5 5% % 6. 6.0 0% % -6 G 6.5 .5% re % at 7 er th .0% an 10 %
0%
Source: Morgan Stanley, Moody’s
HOW MUCH RISK PREMIUM?
As historical loss rates on investment grade portfolios are low, comparing the much higher market-implied loss rates with the historical ones can give us good insight into how much risk-premium investors are getting relative to historical loss rates. In Exhibit 4, we compute the ratio of the expected losses, based on the market pricing (from Exhibit 1) and derivative models, to the historical expected losses. We can think of this as a relative measure of risk premium. On an index-wide basis, it appears that risk premiums currently make up slightly more than half the spread (approximately a 2:1 ratio), and this does not vary significantly for different maturities, which is encouraging. exhibit 4
CDX IG 4
Risk Premium – Market Implied Loss Rates Relative to Historical Loss Rates
5Y Market / Moody's 2.2x
Relative 7Y Market / Moody's 10Y Market / Moody's 2.3x 2.4x
0-3%
1.7x
1.5x
1.2x
3-7%
3.3x
3.1x
3.8x
7-10%
42.6x
9.7x
8.2x
NA
186.8x
43.0x
10-15% Source: Morgan Stanley, Moody’s
For the tranches, however, there is significantly more variation. In the 0-3% tranches, the risk premium declines through time, consistent with the idea that the upfront payments act as subordination. Also, declining risk premiums can be related to the notion that larger upfront payments limit absolute downside.
Please see additional important disclosures at the end of this report.
161
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 18
For mezzanine and senior tranches, we see similar patterns with risk premiums declining through time, which can be reflective of the uncertainty associated with aggregate loss rates for longer maturities. For several of the senior tranches, expected losses based on the Moody’s analysis are so low that the vast majority of spread appears to be risk premium (7-10% and 10-15% for 5-years, and 10-15% for 7-years). Some results from this analysis trouble us. The idea that the equity tranches receive less risk premium than the index, while senior tranches receive more than the index seems strange to us, given that the equity tranches have a levered exposure to the jumpto-default risk in the index, while the higher tranches do not. Additionally, the 3-7% tranche is again problematic, as risk premium actually declines between 5 and 7 years, while it increases from 7 to 10 years. ADJUSTING RISK PREMIUM FOR VOLATILITY
One assumption in the above analysis is that risk premiums scale with the expected default losses for a given asset. This ignores the notion that risk premiums may be influenced by other factors, like price volatility. To take this into consideration, we calculate another measure for each tranche in Exhibit 5. To generate this metric, we calculate the difference between market-implied and historical loss rates to get a sense of the absolute risk premium by tranche. We then divide this result by the absolute risk premium of the underlying index, on a delta-adjusted basis. This delta-adjusted risk premium shows, for example, that most equity tranches are cheaper than the index because of exposure to jump-to-default risk, while mezzanine and senior notes are richer, given their subordination levels. exhibit 5
Index
Risk Premiums Adjusting for Deltas Delta Adjusted Absolute Risk Premium 5-Year 7-Year 1 1
10-Year 1
0-3%
1.22
1.26
0.87
3-7%
0.50
0.63
1.04
7-10%
0.84
0.61
0.66
10-15%
0.75
0.77
0.81
Source: Morgan Stanley
For the 0-3% tranche, risk premium rises marginally between years 5 and 7 and then declines dramatically. For the 7-10% tranche, risk premiums seem high for the 5-year tranche and decline significantly for the 7-year, but then rise marginally for the 10-year, potentially reflective of the increased uncertainty surrounding 10-year default rates and expected losses of 6% for the 10-year maturity. Risk premiums for 3-7% tranches appear to increase with time, although they are relatively flat from 5 to 7 years and then jump sharply for the 10-year maturity. This seems at odds with the cuspy nature of the 7-year 3-7% tranche. While the absolute default risk is not as low as the 5-year tranche, the 7-year prices with a very similar risk premium, despite the fact that it has much greater price volatility. Consider that the 7-year 3-7% has twice the duration of the 5-year, while the 10-year 3-7% is only 50% more volatile than the 7-year.
162
THE STRUCTURED CREDIT BID COULD BE OVERWHELMING
It may be a bit early to have a lot of conviction on the direction of spreads in the 7-year tranche space. We argue that the nature of flows today and in the coming months could be more bullish than bearish for this corner of the market and could push spread tighter for 7-year tranches across the board, despite many of the relative value arguments we have made in this chapter. In particular, in 7-year bespoke structures, activity in mezzanine notes that are more stable than the 3-7% index tranche (say with 5%-like attachment points) may increase to a point where they influence pricing on the standard index tranches, even though this can be a bit like comparing apples and oranges.
Please see additional important disclosures at the end of this report.
163
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 19
Delphi and the Delta Divide
August 12, 2005
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Vishwanath Tirupattur
One of the biggest lessons learned during the auto-stress of the spring was that tranche deltas come in very different flavors, and the market standard ones are designed for small and simple shifts in the underlying index or portfolio, which can actually be irrelevant in many cases. As such, tranche deltas can be misleading for at least three reasons: •
The actual move in the underlying index or portfolio is different from the assumption, including a reshaping of the portfolio distribution.
•
The actual move is similar to the assumption, but the move is big enough to make convexity (gamma) meaningful.
•
There are pricing moves not explained by the models, which can be interpreted as correlation shifts.
Recent news related to the timing of bankruptcy reform and Delphi’s likely negotiating tactics as it related to GM caused some very concentrated market movements in an otherwise quiet August wind-down. This recent jump in Delphi spreads serves as a reminder of the importance of understanding what is behind the tranche deltas, even though the auto-stress in the spring was not very long ago. Given the legacy positions in many bespoke and index tranche transactions, exposure to Delphi in the structured credit is by no means confined to the current CDX HY Index. Delphi was in the benchmark investment grade index through CDX 3 and is also present in many bespoke IG transactions. We estimate there is roughly half a billion dollars of Delphi exposure in rated tranches, although this is likely a small subset of the overall exposure in the structured credit market. In this chapter, we highlight tranche price moves resulting from the action in Delphi and compare realized deltas (from market pricing) and market standard deltas during this tumultuous year to give us better insight into the risks of using standard deltas in environments where idiosyncratic risks dominate. SO MUCH FOR A QUIET AUGUST – DELPHI
Last week, Delphi’s announcement that it was drawing down the majority of its remaining revolver drove benchmark 5 year default swaps 350 bp wider, to 1025 bp. Exhibit 1 provides a summary of the trading levels for Delphi, and some of the index products that contain Delphi, as a constituent from August 3rd through August 8th to capture the full impact of the news. We have also summarized the change in intrinsic value and the portion of this intrinsic change theoretically attributable to Delphi. Interestingly, the actual traded price changes for all of the index products exceeded the Delphi and total intrinsic impacts, highlighting the flows in the index. This is particularly true for CDX HY.
164
Index Observed and Implied Price Moves (bp)
exhibit 1
8/3/2005 675
8/8/2005 1025
Change 350
CDX IG 3 Actual
52.25
55.25
3.00
CDX IG 3 Intrinsic
51.51
54.36
2.85
Delphi 5 Year
Price Impact
-0.12% -0.11%
Intrinsic Change due to Delphi
-0.06%
CDX HV 3 Actual
113.00
127.00
14.00
-0.54%
CDX HV 3 Intrinsic
112.28
123.85
11.57
-0.45%
Intrinsic Change due to Delphi
-0.27%
CDX HY4 Actual
100.50
99.38
(1.13)
CDX HY4 Intrinsic
101.04
100.55
(0.48)
-1.13% -0.48%
Intrinsic Change due to Delphi
-0.09%
Source: Morgan Stanley
TRANCHES – SECOND ORDER EFFECTS
Given the differences in observed price changes versus those theoretically implied by the underlying single-name price changes and our experience in May, it is not surprising that tranches also reacted in a manner somewhat inconsistent with what standard deltas would have predicted. In Exhibit 2 we summarize the tranche price moves for CDX IG 3 along with the implied theoretical price change, based on standard deltas. As Delphi is largely an idiosyncratic event for now, the equity tranche underperformed and the mezzanine outperformed the standard delta-based moves, as we would expect. While it is tempting to point to the isolated impact of Delphi as the source for the mismatch between actual market moves and expectations, we have observed this pattern consistently throughout 2005. CDX 3 Tranche Observed and Implied Price Moves
exhibit 2
Standard Delta Predicted Price Change
7/29/05 Mid 50
8/9/05 Mid 54
Actual Price Change -0.17%
0-3%
38.25%
44.50%
-6.25%
-3.18%
3-7%
102
125
-0.93%
-0.99%
Tranche Index
Source: Morgan Stanley
IMPLYING DELTAS FROM THE 2005 CREDIT ENVIRONMENT
If market standard deltas are not that useful in the risk management of junior tranches, what types of deltas are useful? In Exhibit 3, we show the ratio of the 1 month market implied hedge ratios to correlation model based hedge ratios for the 5 year OTR CDX
Please see additional important disclosures at the end of this report.
165
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 19
index, to illustrate the difference between model implied price sensitivities and those actually observed in the market. We summarize the same values for the 10 year OTR CDX index in Exhibit 4. A ratio greater than 1 indicates actual market price sensitivities that are greater than those predicted by model deltas, and a ratio less than 1 indicates less realized price sensitivity. exhibit 3
Realized Versus Model Deltas – 5Y CDX Tranches
4.0 3.5
0-3%
3-7%
7-10%
10-15%
Feb-05
Mar-05
3.0 2.5 2.0 1.5 1.0 0.5 0.0 Jan-05
May-05
Jun-05
Aug-05
Source: Morgan Stanley
The most immediately noticeable trend is for equity tranches to be upwards of 4 times more volatile than standard deltas would have predicted. Even in times of relative calm (January through March) the observed market based hedge ratio is double what models implied. Mezzanine tranche deltas, while they track closer to the model implied numbers (as illustrated by the lines being closer to 1), tend to be less volatile than expectations but also experience periods when volatility can be greater than expected. The data show similar trends in 10 year with the exception of the 3-7 tranche, which has been universally more volatile than expected of late. This mismatch in the model based deltas and those actually observed is not really all that surprising given the fact that deltas are generally calculated based on some universal spread move assumption, while credit markets in recent months have been dominated by sector or credit specific periods of stress, which can have much different price implications on tranches.
166
exhibit 4
Realized Versus Model Deltas – 10Y CDX Tranches
5.0 4.5
0-3%
3-7%
7-10%
10-15%
Feb-05
Mar-05
4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 Jan-05
May-05
Jun-05
Aug-05
Source: Morgan Stanley
DELTAS OF A DIFFERENT FLAVOR
While the idea of a single delta for a tranche is an attractive concept from a practical perspective, the real world is not quite so simple, especially in the lower parts of capital structures. In Exhibit 5, we show a 1 month backward looking market based delta along with 3 model based delta estimates (given recent market levels) using the following assumptions: all credit spreads move 1 bp, 4 credits widen by 1000 bp, 1 credit widens by 2,500 bp. Conceptually, these assumptions align to a broad market move, a large sector specific move and a single credit becoming distressed. We have selected the autos as the sector to stress for this illustration. As Exhibit 5 illustrates, the underlying spread assumption we use has important implications for the resulting hedge ratio for most tranches. Generally, equity tranches can be viewed as being almost twice as sensitive to concentrated spread moves as they are to broader market moves, while mezzanine and senior tranches are more exposed to broad market moves. While we have discussed this phenomenon many times, the order of magnitude of the differences is worth noting.
Please see additional important disclosures at the end of this report.
167
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 19
exhibit 5
Tranche 5Y 0-3%
The Baskin-Robbins of Delta Broad Market Delta 16.5
Sector Stress Single Credit Market Implied Delta Distress Delta Delta 23.5 28.5 41.0
5Y 3-7%
5.5
8.0
5.9
4.2
5Y 7-10%
1.8
1.5
0.9
1.4
5Y 10-15%
0.9
0.5
0.3
0.8
5Y 15-30%
0.5
0.07
0.03
0.3
Weighted Average Tranche Delta
0.9
1.1
1.1
1.5
10Y 0-3%
5.6
26.3
34.6
16.1
10Y 3-7%
8.6
6.8
4.1
11.7
10Y 7-10%
3.7
3.7
2.3
3.7
10Y 10-15%
2.0
1.1
0.6
1.6
10Y 15-30%
0.9
0.09
0.03
0.9
Weighted Average Tranches Delta
0.8
1.2
1.3
1.3
Source: Morgan Stanley
An interesting implication of this analysis is that the weighted average delta for the liquid tranches is actually greater than 1 based on either of the two concentrated spread assumptions, meaning that the tranches are actually more volatile in aggregate than the entire index (even if we assume the super senior has no price sensitivity at all). One interpretation of this result uses the idea of tranches as options on the portfolio losses. The uniform spread move changes the expected losses but not the uncertainty (or volatility) of the expected losses. The concentrated scenarios change both the expected losses and the volatility, since the losses become more concentrated in a small number of credits. This change in volatility would not affect the price of the underlying index but it would have an impact on the price of the tranches on the index, if we think of them as options on the index losses. The increase in equity and decrease in mezzanine deltas for the concentrated scenarios can be explained as a change in the shape of the market implied loss distribution (in option terms – a change in the skew). WHAT DRIVES TRANCHE PRICING SENSITIVITIES? RECONCILING MODELS AND THE REAL WORLD
Putting technical issues aside for a moment, it seems logical to assume that (under normal conditions) the market actually prices tranches based on some weighted combination of broad based and concentrated spread moves, so the actual observed price sensitivity could be viewed as a combination of these two measures. To the extent the market price moves are more like one than the other, it could indicate the level to which the market is focused on broad spread moves (which we can think of as the price of risk changing) versus concentrated defaults (actual expected defaults changing). In Exhibit 6, we have attempted to quantify this by calculating the weighting implied by market deltas assuming that the actual delta is some combination of these two model implied numbers.
168
The results are somewhat encouraging for 5 and 10 year senior tranches, with the actual price moves appearing to heavily weight broad market risks, consistent with the actual risk in these tranches. For junior tranches in 5 years, the results show how technicals have driven price dynamics to extremes relative to any model derived metrics. The results imply that investors have repriced risk in these tranches, with realized 0-3% price changes indicating an increased focus on idiosyncratic risk in this tranche and a reduced appetite to take that risk. Technicals in 3-7% tranches show the other side of this dynamic, with muted price sensitivity that is heavily focused on broad spread moves – which argues that there is little perceived idiosyncratic risk in this tranche. exhibit 6
Tranche 5Y 0-3%
What Drives Pricing Sensitivity – Broad Market or Single Name Moves? Broad Market Weight -104%
Single Credit Distress Weight 204%
5Y 3-7%
412%
-312%
5Y 7-10%
55%
45%
5Y 10-15%
88%
12%
5Y 15-30%
67%
33%
10Y 0-3%
64%
36%
10Y 3-7%
169%
-69%
10Y 7-10%
97%
3%
10Y 10-15%
70%
30%
10Y 15-30%
99%
1%
Source: Morgan Stanley
CONCLUSION – THE TRANCHE DELTA DIVIDE
Market standard tranche deltas can be misleading for a variety of reasons, most notably because the standard scenario assumptions (small parallel shifts in spreads) may not always be the largest risk factors in an instrument, particularly for junior tranches. The recent events in Delphi serve as another reminder that market standard deltas are not appropriate for big shifts in single-name risk. With continuing volatility in Delphi as we move toward the bankruptcy reform date in October, we encourage junior tranche users to pay close attention to the risks they actually want to hedge.
Please see additional important disclosures at the end of this report.
169
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 20
Reading into Correlation
December 9, 2005
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj
Spreads are the most natural indicators of default risk, and risk-neutral models (which are the bread and butter of derivatives pricing) assume that every ounce of premium above the risk-free rate is compensation for default risk. Yet the relationship of marketimplied default rates and actual default rates can vary over time in significant ways. In investment grade markets, throughout the 35-year period we have tracked, positive risk premium (spreads imply more defaults than actually occur) has always existed. However, in high yield, the volatile nature of actual defaults results in periods of time where risk premium completely disappears and even goes negative (spread does not compensate for default risk). We recognize this has been a hotly debated topic for years and point readers to the December 12, 2002, High Yield Strategy Insights, “Of Cats and Dogs,” for some earlier thoughts. Clearly these relationships affect macro and micro investment decisions, and they have meaningful implications for structured credit investment strategies, which can be viewed as plays on actual default risk versus what is implied by the markets. If watching credit spreads march tighter has been the most popular way to pass time the last couple of years, then blaming structured credit activity for the tightness has been a close second. While many investors these days love to talk the correlation talk, we find meaningful insight into structured credit investing by examining some fundamental relationships in credit markets. RISK PREMIUMS IN CREDIT MARKETS
The question we often come back to is whether spreads do a good job estimating the risk in Corporate America on a forward looking basis. The answer to that question has important implications, not only for the absolute performance of credit markets, but also for the pricing of tranches, since, at their heart, tranches are simply levered plays on default risk. In Exhibits 1 and 2, we perform a simple analysis to test the extent to which spread levels are consistent with realized default risk. In the exhibits, we calculate the ratio of the present value of average spread prevailing in a given year (for an assumed 5-year maturity instrument) to the expected losses derived from Moody’s 5-year cumulative historical default rates and a 40% recovery assumption.
170
exhibit 1
Ratio of Realized Losses to PV of Spread – IG
60%
50%
A BBB
40%
30%
20%
10%
2000
1998
1996
1994
1992
1990
1988
1986
1984
1982
1980
1978
1976
1974
1972
1970
0%
Source: Morgan Stanley, Moody’s, Yield Book
For investment grade (Exhibit 1), we find that losses from default have historically averaged 20% of the total spread value for BBB credits, but this number is fairly volatile, ranging from 50% to 0%. Based on this analysis and given the low likelihood of default, investment grade investors appear to charge a sufficient amount to ensure positive returns (in a very general market-wide sense) through the default cycle. exhibit 2
Ratio of Realized Losses to PV of Spread – HY
180% BB
160%
B
140% 120% 100% 80% 60% 40% 20% 0% 1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
Source: Morgan Stanley, Moody’s, Yield Book
High yield appears to be another matter, with losses for B-rated names having exceeded the present value of spreads in the five-year periods beginning in 1991, 1997 and 1998. On average, losses make up 45% and 88% for BB and B credits, respectively, in the 15year period we examine. While these proportions are much greater on a relative basis, it is worth noting that on an absolute spread basis, the excess premium (adjusted for defaults) historically captured by BB investors exceeds that of both A and BBB investors, while excess premium captured by B investors is actually less than their investment grade brethren. We caution that these results could be a function of the limited data set we use for high yield, which is dominated by the 1990s and early 2000s.
Please see additional important disclosures at the end of this report.
171
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 20
ENTER THE STRUCTURED CREDIT MARKETS
The differences between actual default risk versus premium charged to bear that risk are critical to judging relative value in tranches. Spread changes in underlying portfolios should generally have an impact on tranche pricing, given arbitrage forces. That impact need not be uniform across the capital structure, as we have all learned this year, with volatility in super senior and equity tranches. The market’s perception of the fairness of underlying credit spreads relative to the perceived default risk should be reflected in tranche pricing. Consider the graph of BBB risk in Exhibit 1: To the extent we are in an environment similar to the five years following 1998, junior mezzanine tranches should trade relatively wide, reflecting the increased risk relative to spreads. In a period similar to 1992, we would expect junior mezzanine tranches to trade relatively tight, given the low level of absolute risk, even in the context of tighter spreads. A similar examination of high yield tranches offers further insight. With high yield tranches tending to trade wide to comparable investment grade tranches (on a ratings and implied correlation basis), we shouldn’t fail to consider the possibility that there are times when there is simply not enough spread in high yield to make investors whole, net of defaults. In a world in which spreads are too tight on an absolute basis, tranches we would characterize as out-of-the-money have the significant risk of being in-the-money some of the time. This incremental risk in high yield, related to the fairness of underlying spread, may help explain differences in pricing between investment grade and high yield tranches. TRANCHES AND IMPLIED CORRELATION
While we know that ideas about the meaning of implied correlations have been severely challenged this year, we believe there is still some useful information in implied correlation numbers. However, we have nowhere near the history to make any table-pounding statements about the predictive power of tranche pricing in determining future default risk. We do think it is worth examining current pricing to at least try to get a sense of what the tranche market might be telling us. We examine the implied correlation for the DJ CDX 5 portfolio in Exhibit 3 under two assumptions. The first is the standard assumption that 100% of spread is default risk. The second assumption, designed to show the impact of a small change in the relative pricing of spreads, is that 90% of spread is default risk (for all but the 10 widest names) and the remainder is for other factors. The analysis highlights that both absolute correlation levels and correlation skew are significantly affected by the assumptions we use about the fairness of spreads.
172
Implied Base Correlations Are Flatter
exhibit 3
3%
5-Year Tranches Standard 90% Default Risk 9% 5%
10-Year Tranches Standard 90% Default Risk 8% 5%
7%
25%
15%
14%
9%
10%
34%
18%
24%
16%
15%
45%
17%
36%
24%
7-15% Skew
20%
2%
23%
15%
Source: Morgan Stanley
In the later scenario, we find that correlation skews are flattened dramatically, most notably across the 5-year mezzanine tranches, where the skew is reduced from 20% to 2%. Based on this result, it seems we may be able to glean something from correlation skews about the market’s view of the fairness of single-name spreads, or at least an idea of how much risk premium is perceived to be in the single-name market. A steep correlation curve may indicate a market in which spreads have a reasonable amount of risk premium over the market’s view of expected defaults. Alternatively, a flatter curve could indicate a general view that spreads in the underlying assets are fair to tight relative to default risk. This is consistent with the relative flatness we observe in high yield tranches (given the data in Exhibit 2). Looking across the Atlantic at Itraxx tranches, we find a slightly flatter skew than in CDX, which could indicate relative richness in that market. That result should be considered in the context of an Itraxx 5-year index, which trades at 35 bp, or 30% less than the comparable CDX spread. CONCLUSION
While many have written off implied correlation numbers this year (and this is perhaps appropriate, given how they were being used), we find that the tranched markets can still provide us with some insight about the underlying credit markets or, at least, about tranche investors’ views of those markets. We recognize that small or moderate implied correlation moves can easily be driven by technical flows in the market, but watching for the next big correlation skew change might give investors the edge in identifying inflection points in credit market sentiment.
Please see additional important disclosures at the end of this report.
173
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 21
Tranche Optics – Things Seem Blurry
March 24, 2006
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Vishwanath Tirupattur Primary Analyst: Pinar Onur
The secular has indeed overshadowed the cyclical, as a few important market themes have forced a strong one-way flow in the structured credit markets this year. We wonder now whether it can continue, given the optics of today’s very tight tranche spreads (see “2006 Outlook: The Cyclical vs. the Secular,” December 16, 2005). There are certainly many technical reasons why benchmark tranches are trading where they are, but without the right lenses, we find it difficult to make the argument for them to widen. We focus on both the technical landscape and valuation themes in this chapter, and our key takeaway is that mezzanine index tranches, in particular CDX Series 5 5year 3-7%, seem overvalued by a variety of measures. There are both absolute and relative ways to play this overvaluation theme, but these strategies will only work if the technicals are supportive, at least for a brief period. It has been a very long time since we have recommended buying protection on mezzanine tranches, given the relatively benign default environment and the titanic structured credit bid; however, absolute valuations are now telling us otherwise for certain tranches. MARKET THEMES AND THE TECHNICAL LANDSCAPE
Tight valuations in certain index tranches reflect a technical landscape influenced by some critically important current market themes, which we summarize below:
174
•
Equity tranches remain cheaper on a correlation basis, but not necessarily on a price basis, as a combination of significantly tighter index spreads and continued fears of auto-related default risk reduces appetite for equity on both the buy- and sell-sides.
•
Index mezzanine tranches continue their march through all-time tight levels. We argue that index tranches are being driven by the dealer correlation hedging activity of bespoke flows, but there is some mystery surrounding the impact in the 5-year area, given the 7-year bespoke flow.
•
Bespoke flows continue to be driven by a variety of factors including S&P model changes, a new calendar year, and, of course, our two favorite secular themes: Basel II and FAS 155. There is no change in our view here. Only very tight spreads (like we are seeing now) or a much higher default rate will slow this flow, but there is certainly a back bid.
•
Many blamed the index and CDS roll for the ever tightening CDX 5 tranches, particularly 5 year, with the expectation that protection owners would roll. But with the roll behind us now, we have only seen modest widening of CDX 5 mezzanine tranches.
•
Index tranches serve both as investment vehicles and important dealer hedging tools. When one type of use overshadows the other, it is important to take note. We do not believe that 5-year dealer hedging activity is sustainable, given the sweet spot of bespoke flow, so there is some argument for 5-year tranches to widen. We also point out that there is significant “basis” risk between bespokes and index tranches, and we worry that the dealers who have adopted this approach will experience a mismatch in risks if idiosyncratic events rise. exhibit 1
A Huge Rally – 3-7% and 7-10% 5-Year Tranches
500
3-7% (Left Axis)
7-10% (Right Axis)
180
150
400
120 300 90 200 60 100
0 Oct-03
30
Apr-04
Oct-04
Mar-05
Sep-05
0 Mar-06
Source: Morgan Stanley
TRANCHE VALUATION – CONSIDER OTHER LENSES
Absent market technicals, we see little value in selling 3-7% type 5-year risk at levels near the spread of the underlying index or portfolio. While benchmark 3-7% tranches have not quite reached that point, they are very close. We can recall past debates about whether 7-10% 5-year risk should trade through the index, given negative convexity, but the “money-good” sentiment associated with this tranche, along with ratings, has clearly driven it well through the index. But 3-7% is quite another story. While we still find base correlations useful in some contexts, in today’s market environment the volatile equity tranche correlations and changing relationships among equity tranches of different index series create enough noise in the skew numbers for the immediately senior 3-7% tranches to make them less useful. What else can we look at? We examine two simple alternative relative value metrics. HISTORICAL DEFAULT RATES AND SMALLER PORTFOLIOS
First, for the purist, we revert to our analysis of Moody’s historical data, adjusting for the sampling risk inherent in 125-name portfolios to generate probabilities of various loss scenarios (see Chapter 32). The probability of generating more than 3% losses approaches 6% for a CDX-like portfolio. If we probability-weight these tail events, we get to an expected loss (based on Moody’s history) of about 1%. With pricing on this tranche in the 60 bp neighborhood, market-based expected loss is about 2.7% (PV of 60 bp over 5 years). In our view, there simply does not seem to be enough risk premium imbedded in today’s pricing for a tranche that has a decent amount of default risk, given the attachment point.
Please see additional important disclosures at the end of this report.
175
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 21
exhibit 2
Historical 5-Year IG Loss Rates – The Tail Could Penetrate a 3-7% Tranche
45% Portfolio (125 Names) 40%
Universe (1,000 Names)
35% 30% 25% 20% 15% 10% 5% 0% 0.000.25%
0.250.75%
0.751.25%
1.251.75%
1.752.25%
2.252.75%
2.753.25%
3.25- 3.75% 3.75% or More
Source: Moody’s, Morgan Stanley
SPREAD PER UNIT OF RISK – NOT MUCH ADDITIONAL COMPENSATION
Second, we also consider more market-based metrics, thinking in terms of spread exposure and leverage. We measure leverage adjusted risk by examining the spread on the 3-7% tranche versus the model implied duration on that tranche (using model deltas). In Exhibit 3, we show the historical spread per unit of duration for this tranche and the 7year 7-10% to illustrate that the rally is also significant on a risk adjusted basis. For a tranche that has only 3% of subordination, there is clearly very little additional compensation versus much safer attachment points despite a 2 year longer maturity. exhibit 3
3-7% and 7-10% 5Year Spread Per Unit of Duration
10
5yr 3-7%
7yr 7-10%
9 8 7 6 5 4 3 2 Jan-05
Apr-05
Jun-05
Sep-05
Dec-05
Mar-06
Source: Morgan Stanley
DISPERSION OR CORRELATION – TAILS ARE TOO THIN
So given these lenses, what can we say about tranche valuation? At current valuations, there seems to be ample opportunity to short credit in the most subordinate tranches while being cognizant of the powerful technical forces driving the valuations. One of the key aspects of pricing equity and junior mezzanine risk is the impact of portfolio dispersion, as increased risk in the tail of a portfolio can meaningfully increase risk in the most subordinated tranches.
176
As Greg Peters and team have highlighted in the past, there is little dispersion in spreads today at a time when company fundamentals are beginning to take significantly different paths, often depending on recent equity performance and shareholder pressure on management (see the March 17, 2006, Credit Basis Report, “Credit See-Saw”). The implication of this is that spreads may not fairly reflect actual risk. The second order effect is that risk in equity and junior mezzanine tranches can price rich relative to the actual risk, even if spreads are fair on average and implied correlations look reasonable. POSITIONING TIGHT MEZZANINE TRANCHES
So, to put it all together, how can investors take advantage of this market environment? We have three short-term suggestions: Shorting junior mezzanine outright: At spreads in the 60 bp range, the risk reward seems in favor of buying protection on this tranche, even if only to capture the value of any dispersion increase, much less a widening in underlying spreads. We can see from the current levels on 3-7% tranches, with Series 4 3-7% in 5 years trading about 12 bp wider than Series 5 despite the shorter maturity, the market does attach meaningful value to increased dispersion in mezzanine tranches. Shorting versus delta: While model deltas have proven problematic as mezzanine tranches have outperformed the index, we find that even if we haircut the model deltas in half (our empirical deltas indicate haircuts more like 70-80%) to reflect the strong technicals and significant roll-down in these tranches, we can create positive carry/short spread positions. One example is getting long protection on Series 5 or 6 37% in 5 year versus 2x the index (delta is 4x). As a dispersion play, the index will clearly be less impacted by a widening tail than the 3-7% tranche, all else being equal. We are tempted to suggest this on 7-year 3-7% tranches as well (versus 4x the index), but the technicals here are much harder to lean against. Trading tranches with similar spreads but differing idiosyncratic risk: While tranche spreads appear tight across maturities and the capital structure, risk is not uniform. For example, CDX 6 10-year 7-10% tranches currently trade at 95 bp and have a smaller probability of being penetrated via our Moody’s analysis than the Series 6 5-year 3-7% tranches, which trade 20 bp tighter (0.6% Moody’s expected loss versus 1.0%). Also, the PV of the 95 bp of premium over 10 years is much higher than the 3-7% level for 5 years (7.0% versus 2.7%). From a risk compensation perspective, there is a strong argument for selling CDX 6 7-10% 10-year tranches versus buying CDX 5 3-7% 5-year tranches. Further support comes from roll-down, as well as market technicals, as the 7-10% tranches will continue to have a bespoke bid. A key risk in this trade is potential for better appetite for 5-year risk in a weak market environment than for 10-year risk. TRANCHE OPTICS – TECHNICALS VS. RISK COMPENSATION
Our view on buying select mezzanine protection at these levels is predicated on valuation, but we do not expect significant widening, so investors should view any widening as an opportunity to get out of the trade, unless it is a core short position or a cheap hedge. Based both on default risk views and market technicals, the environment is still very conducive for taking mezzanine risk. Investors seem to be focused on driving spreads to zero on tranches that are “money-good,” but this “theta” (or timedecay play, in derivatives-speak) has gone too far, especially on tranches that are not necessarily “money-good,” at least in our view.
Please see additional important disclosures at the end of this report.
177
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 22
Repeating Correlation 101
March 31, 2006
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Vishwanath Tirupattur Primary Analyst: Pinar Onur
While correlation models have certainly received their share of criticism, a risk-neutral (i.e., market-implied) approach to tranche pricing is both necessary in a derivatives world and very useful in a valuation and relative value sense. Yet, the biggest market themes today, generally cheaper equity tranches combined with much richer mezzanine and senior tranches, pose some difficult challenges in determining relative value among “similar” investment opportunities. And market standard correlation models are not providing that much help. The market today focuses on base correlation as the preferred method for tranche relative value, but it did not start that way (see Chapter 13). The great mezzanine rally that started in late 2003 began to show the weakness of the original “compound” correlation models, as cuspy mezzanine tranches (e.g., 5-year 3-7%) started to price at levels that were either not sensitive to correlation, well below the correlation curve, or where there were multiple correlation solutions for a given spread. Base correlation, in reality a small tweak to the original approach, was the solution – but today we find that it is hiding some important relative value information. We are by no means advocating that the world go back to compound correlation as the sole pricing and relative value tool. In fact, development today is moving in different directions, with a focus on modeling spreads dynamically, which is a shortcoming of both base and compound correlation models, given their focus on current single-name spread levels. Our point is that some important information about tranche pricing gets lost, given the shape of base correlation skew curves and lack of independence from equity tranches, where dispersion certainly exists in market pricing. We go through some details in this chapter and focus on relative value among index tranches.
178
A LITTLE BIT OF CORRELATION HISTORY
In a market that doesn’t have that much history, it is important not to throw any of it away. If you lived through this correlation history, you can skip this brief section, but we are actually amazed how many market participants are unaware of an important pricing and modeling transition that occurred only two years ago. If one were getting a degree in modern-day methods to price synthetic CDOs, the 101 course would have been on compound correlation, and then the remaining classes would have basically ignored this content and focused on common market practices (base correlation, rating agency models, etc.). But if someone asked you some very basic questions about intuition behind high vs. low correlation and pricing of equity vs. non-equity tranches, you would have to dig up your freshman-year class notes, which we do below. exhibit 1
Correlation 101 Gets Lost Today
bp
3-7% (LA)
300
7-10% (LA)
0-3% (RA)
Non-monotonic: sensitivity to correlation changes direction
bp 1,800 1,600
250 1,400 200
1,200
Short Correlation
1,000 150 800 100
600 400
50 200
Long Correlation 0
0 0%
10%
20%
30%
40% 50% 60% Correlation
70%
80%
90% 100%
Note: 5-Year CDX 6 tranches. Source: Morgan Stanley
Compound correlation refers to the correlation implied by tranche pricing, using standard risk-neutral models that allocate default risk based on market spreads. It is indeed the correlation that reflects all of the intuition about low diversity vs. high diversity that we all spoke about years ago (see Chapter 9). Equity tranches are long correlation, while senior tranches are short correlation (see Exhibit 1). Mezzanine tranches are problematic in a compound correlation world because some tranches are not that sensitive to correlation shifts and can have multiple solutions. For example, there are two correlation levels for a 150 bp spread level on the 3-7% tranche (one very low and one very high, see Exhibit 1).
Please see additional important disclosures at the end of this report.
179
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 22
To get these models to be better behaved, the market moved toward using base correlation, which effectively means that a long position in a 3-7% tranche would be modeled as a long position in 0-7% and a short position in 0-3%, both of which are “equity” tranches and thus better behaved. The correlation skew (i.e., the difference of 0-7% and 0-3% base correlation) that came out of this approach is then used to judge relative value. However, the market is actually ignoring the fact that junior mezzanine pricing currently requires extreme events as it relates to expected actual default experience. Note that base correlation for a true equity tranche equals compound correlation (by definition). In a base correlation framework, all tranches behave like equity tranches, meaning that their spreads tighten as correlation skew rises, all else being equal. However, in a true economic sense, equity tranches should prefer higher correlation (less portfolio diversity with respect to defaults) while senior tranches should be afraid of it. While many market participants were quite comfortable with this base correlation method, it tells us very little about what kind of market environment tranche markets are actually pricing in. exhibit 2
Base Correlation Skew – Little Dispersion 5 Year
3-7%
CDX 4 13.9%
CDX 5 14.3%
CDX 6 14.3%
7-10%
9.0%
8.6%
8.9%
10-15%
11.5%
10.9%
11.5%
15-30%
24.1%
23.8%
24.8%
3-7%
CDX 4 11.2%
CDX 5 11.2%
CDX 6 11.5%
7 Year
7-10%
8.5%
8.4%
8.8%
10-15%
11.7%
10.8%
11.2%
15-30%
26.6%
26.1%
27.0%
3-7%
CDX 4 5.4%
CDX 5 4.3%
CDX 6 3.5%
10 Year
7-10%
9.3%
9.1%
9.0%
10-15%
10.6%
11.1%
11.2%
15-30%
28.2%
28.3%
28.6%
Source: Morgan Stanley
180
BASE CORRELATION LENS – THE WORLD IS FAIRLY PRICED
With the history lesson behind us, why is it important to reintroduce an old concept? In Exhibit 2 we summarize the current implied base correlation curves for the CDX IG series 4, 5 and 6 tranches. What jumps out to us is the similarity among the base correlation curves and their skew. So from the perspective of current models, it looks like tranches are priced reasonably fairly, relative to one another. While equity correlations have recently become more consistent across the series, any dispersion in equity correlations (like we saw just a few weeks ago) would have the follow-on effect of making skew comparisons across the index series that much more difficult to interpret. However, regardless of what base correlation says, our intuition, portfolio composition differences and optical differences in tranche pricing do seem to suggest there is value to be found across capital structures, maturities and/or index series. COMPOUND CORRELATION LENS – THERE IS SOME RELATIVE VALUE
Given the homogeneity in base correlation curves, we next examine the implied compound correlations for the various on-the-run tranches in CDX IG (see Exhibit 3). While, there is a good amount of homogeneity among the CDX series here as well, a few things do jump out at us. exhibit 3
Compound Correlation – More Dispersion 5 Year
0-3%
CDX 4 8.54%
CDX 5 9.61%
CDX 6 9.11%
3-7%
4.20%
3.20%
1.86%
7-10%
10.53%
10.50%
8.83%
10-15%
17.47%
17.39%
15.22%
15-30%
29.54%
28.98%
27.45%
0-3%
CDX 4 5.39%
CDX 5 5.49%
CDX 6 5.13%
3-7%
80.29%
84.31%
78.29%
7-10%
6.34%
5.56%
4.47%
10-15%
12.44%
12.39%
11.45%
15-30%
23.34%
22.22%
21.35%
0-3%
CDX 4 5.23%
CDX 5 6.42%
CDX 6 5.74%
3-7%
13.79%
16.99%
13.99%
N/A
N/A
N/A
7 Year
10 Year
7-10% 10-15%
7.87%
6.75%
5.52%
15-30%
17.82%
16.89%
15.72%
Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
181
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 22
In 5 year tranches, 7-10% and 10-15% in Series 6 look to be pricing a bit rich versus the other series. The market is currently pricing the additional 6 months of risk on a marginally higher quality portfolio (versus 5) at about 2 bp (net of bid/offer or 3.25 mid/mid) for 7-10%, which doesn’t sound like all that much. However, given absolute spreads, this does represents about 13% of the total Series 6 spread present value in this tranche. The question that remains is whether this increase in risk is enough despite how out of the money these tranches actually are and the relative lack of event driven risk in this portfolio. We point out that each compound correlation point on this tranche is worth about 2 bp. In 7 year space 7-10% tranches look to be very close to fair value on a base correlation basis, but compound correlation indicates that Series 6 is richer than Series 5, which in turn trades rich to Series 4. While this trend may be justifiable given declining risk aversion associated with less dispersed and less event driven portfolios in the newer Series, base correlation misses this trend. A compound correlation point in this tranche is worth about 7 bp. In the 10 year space, 3-7% tranches are clearly equity like (see Exhibit 4), and there appears to be some interesting relative value among CDX 4, 5, and 6. 3-7% in CDX 5 prices the richest (3 correlation points higher with each correlation point worth about 8 bp) despite the fact that the underlying portfolio is not that different from Series 6 (Series 5 has American Axle and Albertson’s). This same theme is true in 0-3% tranches where each correlation point is worth about 35 bp. The smaller correlation difference and larger absolute spreads in 0-3% suggest that the 3-7% price differential is larger relative to the risk and spread embedded in the tranches. We note that using base correlation skews, these pricing differences are lost, given the lack of independence from 0-3%. exhibit 4
Names Can be Confusing – 3-7% Changes Character
bp 5 yr
900
7 yr
10yr
800 700 600 500 400 300 200 100 0 0%
Source: Morgan Stanley
182
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
CONCLUSION – LOST INTUITION AND RELATIVE VALUE
Our point with this chapter is that a market as large and deep as the structured credit one should not throw away metrics that have intuitive appeal. Base correlation is an important tool but it is not by any means all-powerful. The fact that the many index tranche opportunities look fairly priced on a base correlation basis shows that too many investors are choosing to use only this one lens. Compound correlation has issues as well, yet given that it is based on exactly the same market inputs as base correlation and has more tranche-specific independence, it is quite useful.
Please see additional important disclosures at the end of this report.
183
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 23
Can Spreads or Correlation Go to Zero?
April 7, 2006
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Vishwanath Tirupattur Primary Analyst: Pinar Onur
While somewhat jocular, a question about whether tranche spreads or correlation could go to zero is both a popular query today and a topic that actually highlights the relationship of economic cycles and default risk. Although a bit of widening has materialized in the most junior mezzanine tranches of late, the theme of generally tighter senior tranche spreads is loud and clear. This is both a roll down and a correlation phenomenon (see Exhibit 1), with the net effect being single- and lowdouble-digit spreads on many senior tranches. Can this trend continue? We are not suggesting that spreads can go to zero on any kind of derivative instruments, given that investors demand some amount of risk premium for being involved. However, low spreads on underlying credits today suggest that many senior tranches will only get touched when we have much higher default correlation, i.e., when there is some systemic force (like economic cycles) that causes credits to default together. If this does not exist (i.e., correlation itself is zero or close to it), then defaults are truly independent of one another, and default probabilities (on underlying credits) are not high enough to pose any real risk of loss in certain senior tranches. So a question about whether spreads can go to zero is in some sense a question about whether correlation can go to zero, at least for senior tranches. For junior mezzanine tranches, the zero correlation point is actually a positive spread, which is a good indication of a floor on spreads, at least in a risk-neutral sense. Why is any of this relevant? With the tremendous rally in most senior and many mezzanine tranches, investors wonder where the bottom is. For some tranches, it is indeed some number near zero, given today’s credit environment, but for others it is not, which can be helpful in identifying points of entry for shorting tranches, a recent example being 5-year 3-7% tranches (see Chapter 21).
184
WHAT HAS DRIVEN SENIOR TRANCHE RETURNS?
For the most part, senior tranches have been a one-way trade over the past three years. The biggest drivers have not been the obvious factors, namely tighter underlying markets and the actual carry on the tranches. By a large margin, the most important contributors to return have been roll down (i.e., time decay) and correlation skew. The roll down return comes from the fact that default rates have been far lower for the onthe-run IG indexes than implied by the market. As such, the “out-of-the-money” nature of senior tranches gets even stronger as time passes. The correlation skew return comes from the markets’ belief that it will continue, pushing tranche spreads even tighter relative to what risk-neutral models say. This skew is similar to what we see in options markets. For 5-year senior tranches, the combination of these two factors is largely the reason for positive total returns over the past 2 ½ years (see Exhibit 1). In the 10-year space, roll down and correlation skew more than compensate for the negative contribution of other factors. exhibit 1 Tranche 5 Year
Roll Down
What Has Driven Senior Tranche Returns? Correlation Skew
Other Factors
Total
7-10%
2.57%
3.38%
0.79%
6.74%
10-15%
1.34%
1.75%
-0.11%
2.98%
7-10%
10.42%
5.38%
-9.87%
5.93%
10-15%
3.00%
5.72%
-3.40%
5.32%
15-30%
1.12%
1.09%
-0.60%
1.61%
10 Year
Note: 5 year returns from 11/03 through 3/06. 10 year returns from 12/03 to 3/06. Other factors include underlying spread moves, curve shifts, parallel correlation shifts, spread dispersion, and carry. Returns are from the on-the-run series of indices. Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
185
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 23
HOW CAN CORRELATION GO TO ZERO?
When we speak of correlation going to zero on senior tranches, we are referring to “compound” correlation, or the implied correlation that comes out of the models, not base correlation skew (see Chapter 22). Low levels of implied correlation can be interpreted broadly as indicating an environment in which defaults are driven much more by idiosyncratic events than by a recessionary environment. Higher correlations (on a relative basis) could be interpreted as indicating an environment where a troubled economy has a sufficient impact on credit markets to drive default rates cyclically higher. In Exhibit 2, the upward sloping line anchored at zero spreads reflects the notion that as correlated default risk falls, the truly independent default propensity of the underlying portfolio (implied by spreads) is not big enough to ever cause a loss in a given senior tranche (5-year 7-10% in this example). exhibit 2
What Happens When Correlation Goes to Zero?
bp
3-7% (LA)
7-10% (LA)
0-3% (RA)
bp 1,800
300
1,600 250 1,400 1,200
200
1,000 150 800
0% Correlation Above 0bp Spread Level
600
100
400 50
0% Correlation Equals 0bp Spread Level
200
0
0 0%
10%
20%
30%
40% 50% 60% Correlation
70%
80%
90% 100%
Note: 5-Year CDX 6 tranches. Source: Morgan Stanley
The absolute risk can cloud these interpretations: there will always be significant risk premium for first loss risk and very senior risk is likely to trade in line with similarly rated competing assets. The places where the markets’ interpretation of the future would be the most evident is in the cuspy parts of the capital structure (i.e. junior mezzanine tranches). When mezzanine prices trade off the implied correlation chart (i.e. tighter than implied correlation models can achieve), this is a strong indication that the market is pricing a healthy credit environment broadly within the context of current spread levels.
186
HOW HIGH CAN CORRELATION GO?
While we have been discussing the blue sky scenario, a more cyclically correlated downturn in credit markets has significant tranche implications. Although correlation sensitivities will tell you that equity tranches benefit from correlation rising and senior tranches are hurt by rising correlation, a dose of common sense helps put bookends around how divergent paths can become. Most credit investors will agree with the statement that a default is likely attributable to some combination of macro events in the economy, micro events in the industry and company-specific events. While one of these could be the dominant driver, it is unlikely that a company defaults solely because of the macro economy. Yet, this is the cause that is most likely to create high correlation among companies in different industries. Similar arguments can be made for the condition of a given industry. As an extreme example in the airline space, consider that while Delta, Continental, Northwest and United defaulted, AMR along with most non-legacy carriers did not. So this raises an important question: Do implied correlations north of say 50% have any economic interpretation, beyond a depression-like scenario? Our interpretation is that constraining ourselves to reasonable correlations numbers (i.e., lower than 50%) makes the job of understanding pricing that much easier. Consider the differences in the shape of the 3-7% curve (Exhibit 2) for correlation less than 50% vs. correlation greater than 50%. SO HOW TIGHT CAN TRANCHES GO?
To get a sense for the extremes that may be possible in tranche pricing, we examined the implied spread on various mezzanine and senior tranches in a zero correlation environment (i.e., the healthiest credit market, see Exhibit 3). For the more senior tranches, a zero-correlation world implies no default risk and thus zero spreads, which makes the argument for buying protection on these tranches in this environment very difficult. In fact, we would argue 7-10% in 7 years and even 10-15% in 10 years are not priced tight enough given this view of the world, although the argument for having more risk premiums for longer maturities (relative to 5 years) is a good one, given increased uncertainty as to the economic environment over time. Junior mezzanine tranches are interesting in this context. Zero correlation spreads in the 30 to 60 bp range for 5 year 3-7% were helpful from the perspective of calling for a widening a short while back (again, see Chapter 21). But we see that 3-7% trades through zero-correlation points in 7-years (where it is still mezzanine-like at these levels). Why is that not necessarily a signal to buy protection? Market technicals are a good reason, as 7-year bespoke activity forces many dealers to hedge using 7-year 37% even though there is significant basis risk between it and the actual bespoke in question (basis risk here refers both to attachment points and portfolio composition). The same argument can be made for 7-10% in 10 years, another sweet spot in the bespoke market.
Please see additional important disclosures at the end of this report.
187
188 0
10-15%
Source: Morgan Stanley
10-15%
2
107
0
10-15%
10 Yr 7-10%
1
7-10%
270
0
7 Yr 3-7%
25
7-10%
2
103
0
1
255
0
0
32
5
162
0
4
339
0
0
62
Spread at 0% Correlation CDX 4 CDX 5 CDX 6
53
103
17
39
244
6
9
77
49
94
17
33
220
7
12
78
53
106
21
40
256
9
18
102
Current Spread (Mid) CDX 4 CDX 5 CDX 6
Where Is the Floor for Spreads?
51
-4
17
38
-26
6
9
52
47
-9
17
32
-35
7
12
46
48
-56
21
36
-83
8
17
40
Difference (Current Spr. – Spr. at 0% Corr.) CDX 4 CDX 5 CDX 6
chapter 23
5 Yr 3-7%
Series
exhibit 3
Structured Credit Insights – Instruments, Valuation and Strategies
CONCLUSION
While the theoretical exercise of understanding whether spreads and/or correlation can go to zero is interesting (you may not agree), the real world implication is finding floors on tranches, as everything threatens to roll down and eventually go to zero. The zero correlation approach is one method, but a key shortcoming in today’s correlation models (and any interpretation of them) is that underlying spreads are dynamic but not modeled as such. So a market that drives senior tranche spreads to near zero is one that is implying low levels of spread volatility in the context of tight absolute spreads.
Please see additional important disclosures at the end of this report.
189
Section C
Investment Strategies
Please see additional important disclosures at the end of this report.
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 24
The Long and Short of It
August 22, 2003
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Anisha Ambardar We believe the credit market is near an inflection point from which spreads and volatility will move either higher or lower together. Based on that view of the credit world, we think scenarios where spreads slowly continue to grind in or jump out dramatically are both worthy of consideration. Traditional trading strategies generally allow investors to benefit from one of these outcomes but not both. Investors who go long yieldy paper will benefit in a low volatility/spread regime but will be the ones who suffer most in a world of heightened volatility/spread levels. Ideally, we would like a position that generates positive carry in a low-volatility world while providing short credit spread exposure in a highvolatility world. How can investors get around this fundamental trade-off? TRANCHING THE TRADE-OFF
CDO investors and other users of tranched credit have long understood the leverage inherent in tranched credit positions. This leverage also creates differing convexity characteristics depending on the level of subordination (see Chapter 10). With the advent of a traded tranched TRAC-X market, investors have the ability to customize exposure to spreads and convexity. In Exhibit 1, we illustrate two sets of long/short combinations in tranched TRAC-X. Both strategies are structured to have offsetting price changes for very small moves in TRAC-X spreads and both have the same short notional position in the 10-15% tranche. Exhibit 2 illustrates the six-month forward P/L for both strategies, as generated in our fair value model. The first strategy offers a moderate amount of leverage and convexity. While this strategy offers positive carry, the exposure is directionally equivalent to a short position in widening credit spread scenarios and a long position in tightening scenarios. The net carry of the combined positions is 65 bp.
192
exhibit 1
Summary of Long/Short Tranched TRAC-X Strategies Long Position
Short Position
Strategy 1 Tranche
TRAC-X
10-15%
Notional
100,000,000
50,000,000
Recent Spread (bp)
73
65
Net Carry (bp)*
65
Strategy 2 Tranche
0-3%
10-15%
Notional
11,000,000
50,000,000
500 Plus 51.5% Upfront
65
Recent Spread (bp) Net Carry (bp)*
284
*As a percentage of the long position notional amount, adjusted to reflect financing of any upfront payments. Source: Morgan Stanley
The second strategy offers a higher degree of leverage. This leverage also generates much greater convexity, resulting in a package with a higher overall sensitivity to spread levels, creating the more u-shaped profile. The net carry of the combined positions is 284 bp of the long position notional amount. exhibit 2
6,000,000
Six-Month Forward P/L
5,000,000
Strategy 2
Strategy 1
4,000,000 3,000,000 2,000,000 1,000,000 0 -1,000,000 -2,000,000 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200 TRAC-X Spread
Source: Morgan Stanley
NOT QUITE A FREE LUNCH
From Exhibit 2, one can see that these strategies perform well for most large widening and tightening moves in the underlying TRAC-X spread while resulting in only moderately negative results for minor changes in spread. The carry and spread sensitivity that is created may be attractive, but we have also created an exposure to correlation and modified the nature of the outright default risk.
Please see additional important disclosures at the end of this report.
193
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 24
IMPLIED CORRELATION MATTERS
The valuations in Exhibit 2 are performed using a constant correlation assumption and the value of these positions is clearly sensitive to that assumption. As we have written in the past (see Chapter 9), we view implied correlation as a measure of relative value in the pricing of the tranched credit market, akin to implied volatility in option markets. Therefore, these positions inherently introduce exposure to changes in the prevailing relative value regime in the market for tranched credit. Simply put, both of these strategies are long correlation. More specifically, both strategies have exposure to declining correlation in the 10-15% tranche, which would imply that the tranche trades at tighter spreads, with all else being equal. The second strategy also has exposure to declining correlation in the 0-3% tranche, which would imply that the position trades at wider spreads, all else being equal. Exhibits 3 and 4 illustrate the impact on both strategies of a shift in correlation to 15%, as well as 25%, for both tranches. Recent market implied correlations have been approximately 18% and 20% for the 0-3% and the 10-15% tranche, respectively. One can clearly see that the price paid for the levered exposure to spreads is a significant sensitivity to correlation as well. AND DON’T FORGET DEFAULT RISK
These strategies generally benefit from a widening in TRAC-X spreads but in the extreme scenario of default, they will suffer. In both strategies, investors have first loss exposure to the names in the portfolio, while the offsetting short position has a significant level of subordination. This subordination in the short position will result in a net loss for both strategies in the case of a default in one of the underlying credits. In the first strategy, the nominal default exposure to each name in TRAC-X is approximately $1 million, while in the second strategy, the nominal exposure is approximately $3.7 million. The second strategy has roughly four times the default exposure to each credit, albeit with an identical short position. This increase in nominal default exposure is another aspect of the leverage built into the second strategy.
194
exhibit 3
Strategy 1: P/L Sensitivity to Correlation
6,000,000
15% Correlation
Base Case Correlation**
25% Correlation
5,000,000 4,000,000 3,000,000 2,000,000 1,000,000 0 -1,000,000 -2,000,000 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200 TRAC-X Spread
**Base Case Correlation is 20% for 10-15% tranche. Source: Morgan Stanley
exhibit 4
Strategy 2: P/L Sensitivity to Correlation
6,000,000
15% Correlation
Base Case Correlation**
25% Correlation
5,000,000 4,000,000 3,000,000 2,000,000 1,000,000 0 -1,000,000 -2,000,000 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200 TRAC-X Spread
**Base Case Correlation is 18% for 0-3% and 20% for 10-15% tranche. Source: Morgan Stanley
CONCLUSION
While there are risks investors must bear in order to be involved in the market for tranched credit risk, our view is that the ability to customize exposure to spread movements and convexity may well be worth it. However, investors should carefully consider the correlation and default risk they assume with these strategies. We have provided two examples of trading strategies and encourage investors to explore the potential of this market further.
Please see additional important disclosures at the end of this report.
195
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 25
Getting Long Spread Decompression
October 24, 2003
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Anisha Ambardar We have written much about strategies to short the market using tranches, given the interesting convexity and carry opportunities in today’s market environment. From an underlying credit perspective, the focus has largely been on relatively simple and standard portfolios (TRAC-X, 1 for example). Yet, if we dig deeper into the investment grade market, we find even more interesting short credit/long convexity opportunities. It is easy to construct well diversified portfolios that are more compressed and thus trade significantly tighter than TRAC-X. We argue that such portfolios are good short positioning opportunities, in the context of a larger investment strategy. Another factor that we have often glossed over is the inherent performance difference in shorting portfolios versus shorting tranches when markets do not move in tandem, as recent widening in Ford and GM has illustrated. In a straight credit portfolio, the impact of a decompression in spreads is fairly straightforward, and can be summarized by looking at spread averages. In a tranche, a decompression can have a very dramatic impact, given that portfolio losses are not distributed evenly. Our simple message is to hunt for such opportunities and get long the asymmetric performance. A PORTFOLIO OF THE QUESTIONABLY TIGHT
A 60 handle on TRAC-X indices certainly raises lots of eyebrows and continues to get short sellers excited, but these benchmark indices may be too kind a measure of the relative richness of many credits in the market. There are at least 50 reasonably liquid US investment grade credits with default swap premiums 30 bp or tighter on the offer side. As an illustrative and simple example, we built a diversified portfolio of 115 US credits, where protection on average is offered with a 40 handle. When comparing the portfolio to the TRAC-X series, we find that it is significantly more compressed (only 10 credits trade wider than 60 bp) and is higher rated with marginally higher diversity (see Exhibit 1). From an investment strategy perspective, the portfolio is both a cheaper short and a better opportunity to position for a decompression in spreads.
1
196
SM
Now DJ TRAC-X .
exhibit 1
No. of Credits (Total)
Get Long Spread Decompression, at Cheaper Levels
Example Portfolio 115
TRAC-X NA Series 2 100
No. of Credits (Spread Buckets) 0-30
25
22
31-60
80
41
61-90
10
18
91+
0
19
Average Spread
40
65
Moody’s WARF
232
315
64
58
Moody’s Div. Score Source: Morgan Stanley
We work with this portfolio to illustrate some short credit/long convexity positions. Our main point is that, for credit pickers, this portfolio is one of numerous potential baskets that is well suited for this exercise. Liquidity is the key trade-off in using such customized portfolios for trading strategies versus market benchmarks like TRAC-X. However, for investors with a strong conviction about specific names or portfolios, we believe this is a worthy trade-off. POSITIONING FOR SPREAD DECOMPRESSION
The average beta for all credits in the market must equal one, by definition. But not all credits have a beta equal to one, which is what makes credit markets interesting. Given the incredible amount of spread compression we have seen over the past year, the market is much more homogenous today. But it won’t stay this homogenous forever. Credit- and sector-specific issues, along with general market moves, are catalysts for decompression. How should one position for spread decompression? In a traditional credit portfolio, the simple trade is to underweight high-beta names, given that their incremental impact on performance is directly related to the size of the position. One can think of this relationship as being linear. In tranched products, spread decompression is not a linear phenomenon. The widening of a handful of credits can have a dramatic impact on subordinate and/or “thin” tranches. Why is this so? It is the essence of a capital structure, where the distribution of losses is prioritized. As the likelihood of losses in a portfolio grows, the tranches with the greatest direct exposure to these losses will be impacted more. A short credit position in a reasonably subordinate tranche is a long spread decompression trade.
Please see additional important disclosures at the end of this report.
197
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 25
THREE TYPES OF TRANCHES, TWO TYPES OF DECOMPRESSION
Using the portfolio described in Exhibit 1, we focus on the performance characteristics of three tranches over two types of spread widening and decompression scenarios. The first scenario is a general market widening, where spreads widen equally on a percentage basis (i.e., if a 50 bp credit moves 5 bp, a 100 bp credit would move 10 bp). The second scenario is a sector blowup, where a handful of credits widen out dramatically relative to the rest of the market. We propose three tranche strategies to position for a market widening and spread decompression (see Exhibit 2). The notional amounts are such that the net negative carry (from the purchase of protection) for the three positions is equivalent. Tranche Strategies
exhibit 2 Notional (MM) $163.3
Tranche Portfolio
Premium Net Carry (MM) (Offer Side) (Notional x Prem) 45 $(0.75)
Correlation N/A
4-5%
$26.0
283
$(0.75)
5-10%
$50.0
147
$(0.75)
18% 30%
5-15%
$76.6
96
$(0.75)
30%
Source: Morgan Stanley
MARKET-WIDE DECOMPRESSION – 5-15% TRANCHE WINS
Under a market-wide spread decompression scenario, all of the short credit tranche positions outperform the underlying portfolio protection, assuming a fixed correlation (see Exhibit 3). The outperformance of tranche protection relative to the portfolio protection can be explained by the inherent leverage in these positions. The 5-15% tranche is the winner, though, when spreads widen in this fashion. This “thickest” tranche outperforms because, in this scenario, default risk is increasing in a uniform manner across the entire portfolio. exhibit 3
Market-Wide Spread Decompression – “Thicker” Tranche Wins
$ 25,000,000 Portfolio
4-5%
5-10%
5-15%
20,000,000
15,000,000
10,000,000
5,000,000
-
(5,000,000) 20
40
60
80
100
120
140
Average Portfolio Spread
Source: Morgan Stanley
198
160
180
200
SECTOR-SPECIFIC DECOMPRESSION – 4-5% TRANCHE WINS
We simulate a sector specific blowup in spreads by widening out 15 credits in the 115name portfolio (the performance is illustrated in Exhibit 4). The 4-5% tranche is the winner in this case, which seems intuitive given the relative “thinness” and lower attachment point of the tranche and resulting increased exposure to the 15 “risky” names. In the context of a larger investment strategy, we continue to recommend short credit positions in tranches where convexity is attractive. For credit pickers, positioning for spread decompression is an added dimension to consider. exhibit 4
Sector-Specific Spread Decompression – “Thinner” More Subordinate Tranche Wins
$ 25,000,000 Portfolio
4-5%
5-10%
5-15%
20,000,000
15,000,000
10,000,000
5,000,000
-
(5,000,000) 40
60
80
100
120
140
160
180
200
Average Portfolio Spread
Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
199
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 26
Throwing a Curve at Correlation
November 21, 2003
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Anisha Ambardar Correlation trading was certainly the new, new thing in credit markets during 2003, but as we approach the end of this innovative and historically significant year, credit curves are now getting a lot of the attention. Investors should be excited about the development of a “pure” term structure for credit, and the implication of this development goes beyond the single-name world. In the context of correlation markets, the development of credit curves influences both valuation and trading strategies. Much of the initial focus in this area was on understanding valuations at an absolute level and then looking at relative value among tranches. Yet these are just two sides of the three-dimensional correlation puzzle. Considering both the shape of credit curves and the rationale of getting long or short to a particular term are critical tasks in managing a structured credit portfolio. We have argued that default rates and credit spreads are very economically cyclical. Credit curves do not fully reflect this behavior, in our view. For investors who want to be outright long on a levered basis, shorter maturities seem attractive to us, given the typical length of economic expansionary cycles and current market pricing. In more balanced strategies (levered long with short positions), we favor getting short credit through longer-maturity tranches given pricing, the uncertainty associated with cycles turning and the natural spread duration decay in such a position. exhibit 1
How Long Will It Last? US Expansionary Periods
Economic Cycles 1854-2001 (32 Cycles)
Length of Expansionary Period 3.2 Years
1945-2001 (Post WWII, 6 Cycles)
4.8 Years
Feb 1961 – Dec 1969
8.8 Years
Mar 1991 – Mar 2001
10.0 Years
Source: Morgan Stanley, National Bureau of Economic Research
SPREADS, DEFAULT RATES AND ECONOMIC CYCLES
Any levered long credit position today has an implicit view on the length of this expansionary period, given that default rates are very economically cyclical. The average length of expansionary periods over the past 150 years is 3.2 years, but since World War II the average length increased to 4.8 years (based on NBER data; see Exhibit 1). The expansionary periods in the 1960s and 1990s (which lasted 9 and 10 years, respectively) certainly skew this data. With the end of the last recession officially sanctioned as November 2001, we are now two years into the current expansionary period, so the natural question to ask is, how long will it last? Credit
200
curves can give us some indication, and currently the market is telling us that incremental risk in the 3-5-year area is high (steep curve), but the risk between years 5 and 10 is low (flat curve). How relevant is this to levered investors? A LEVERED LONG ONLY STRATEGY
In a perfect world, an investor would structure a levered long credit investment to coincide precisely with the expansionary period, and watch it expire before markets start pricing in trouble. But we do not live in a perfect world, and the risk of running over the cycle should not be ignored, as the resulting performance can be catastrophic. Let’s focus on an example. In Exhibit 2 we compare the pricing of three mezzanine positions: 2-5% and 3-7% tranches on the benchmark 5-year TRAC-X II index, and a 2-5% tranche of a hypothetical shorter version of TRAC-X II (3½ years). Three- and Five-Year Mezzanine Strategies – How Does the Pricing Compare?
exhibit 2
Maturity 3.5
Premium (Mid) 423
Implied Correlation 28%
Spread DV01 38.1
2-5%
5.3
622
31%
52.2
3-7%
5.3
325
*
43.3
Tranche 2-5%
*This tranche is correlation invariant and trades roughly 110 bp rich to model valuation at 28% implied correlation. Source: Morgan Stanley
How can we compare the pricing differences between the tranches? Correlation is one component, and we see the 3½-year tranche is slightly cheaper on a correlation basis (28% vs. 31%). As a caveat, a 3½-year version of TRAC-X II is not standard, and therefore does not trade in the market, but tranches from older (original 5-year) synthetic CDOs do trade. Our 28% correlation assumption comes from the pricing we have seen for such tranches. Apart from correlation, we attribute the bulk of the remaining pricing difference among the tranches to the cumulative effect of default risk and the shape of the credit curve, which is steep between 3 and 5 years (about 24 bp). Today, the 2-5% 3 ½-year tranche and the 3-7% 5-year tranche have similar DV01’s and similar fair spreads (as valued with a flat correlation term structure). Yet, actual market pricing and forward exposures are quite different. What is the right trade? It depends, in part, on one’s view of the economic cycles. AVOID RUNNING OVER THE CYCLE
Going back to our fear of running over the cycle, suppose spreads start widening and default risk starts rising in three years. Intuitively, we are better off holding the 3½year paper versus the 5-year investment under this circumstance, if the spread widening and default risk is much worse than implied by the credit curve. Furthermore, unlike bonds, the spread duration behavior of tranches is not necessarily linear over time. In fact, the DV01 of the shorter tranche decays at a faster pace than the longerdated tranches (see Exhibit 3), which is helpful if the investment runs over the cycle.
Please see additional important disclosures at the end of this report.
201
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 26
exhibit 3
DV01 Decays at Different Speeds
Spread DV01 80.00
2-5% 3.5 Year
0-3% 5.5 Year
2-5% 5.5 Year
3-7% 5.5 Year
60.00
40.00
20.00
0.00 0.0
0.5
1.0
1.5
2.0
2.5 3.0 3.5 Time (Years)
4.0
4.5
5.0
5.5
Source: Morgan Stanley
As such, for investors who are looking to set up leveraged long positions (without downside protection), we favor less subordinated shorter-maturity opportunities, given the big differences in residual risk for an early end to the expansionary cycle. An investor’s appetite for riskier underlying portfolios should be healthier in this scenario as well. exhibit 4
Tranched Trading Strategies – Short-Dated Short or Long-Dated Short? Long Position
Short Position
Package
Package 1 Tranche
2-5%
7-10%
Maturity
3/20/2007
3/20/2009
Notional
$30MM
$51MM
DV01
38.1
22.5
0
Spread
423
115
228*
Package 2 Tranche
2-5%
10-15%
Maturity
3/20/2007
3/20/2014
Notional
$30MM
$37MM
DV01
38.1
31.1
0
Spread
423
115
282*
*Package net spread relative to long position notional amount. Source: Morgan Stanley
LONG/SHORT TRANCHE TRADING STRATEGIES
We have discussed in past research the benefits of balancing long credit positions in subordinate tranches with short positions in more senior tranches (see Chapter 24). These trading strategies have traditionally been implemented with similar maturities (usually on a delta-neutral basis). Yet, a short credit position in a longer-dated tranche may be a better hedge against an unexpected shift in economic cycles.
202
As an example, we take the levered long position described before (2-5%, 3½-year tranche) and delta hedge it by buying 5-year protection on the 7-10% or 10-year protection on the 10-15% tranche of TRAC-X II. Both trades are significantly positive carry (228 bp and 282 bp; see Exhibit 4) and are DV01 neutral at inception. The two packages have different “residual” risks to spread movements over time, which we find interesting strategically. In the first package, the DV01 initially rises and then falls back to neutral over the 3½-year term. However, in the second package (with the 10-year tranche) the DV01 decays significantly over time. Effectively, this second package is a positive carry and DV01-neutral trade today, but becomes short DV01 over time, as the 3½-year tranche decays much more rapidly than the short 10year position. Note that buying 10-year protection expresses a steepening view, but we find support for this fundamentally (see Chapter 24). exhibit 5
Where Do You End Up? DV01 Decay of Long/Short Strategies
Spread DV01 5.00
0.00
-5.00
-10.00
Short 7-10% 5 Year
-15.00
Short 10-15% 10 Year -20.00 0.0
0.5
1.0
1.5 2.0 Time (Years)
2.5
3.0
3.5
Source: Morgan Stanley
ALL ABOUT HEDGING ECONOMIC CYCLES
The rationale for putting on a DV01-neutral trade that decays over time is really all about hedging economic cycles. The largest risks to levered long strategies are defaults and rising spreads. Idiosyncratic or not, defaults are bad at any time during the trade, and the only way to protect against them is subordination (which 2-5% tranches have) and single-name delta hedging, which can be funded by the positive carry. Rising spreads, on the other hand, are only bad when spread duration is high. We have shown both long DV01 strategies that decay rapidly to zero and DV01-neutral strategies that become short DV01 over time. Our fundamental message: use the developing term structure to position levered long or long/short tranches for the cyclical nature of our markets. Otherwise, unexpected economic cycles could ruin a great batting average.
Please see additional important disclosures at the end of this report.
203
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 27
Less Risk, but Longer Tails for Tranches
March 5, 2004
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Anisha Ambardar Recently, we argued that equity and credit markets have diverged fairly significantly this year. Implied equity volatility has fallen in a rallying stock market, while CDS premiums are wider. The Merton-type models are forecasting less default risk on average, but the dispersion of this equity-implied default risk is much greater than in the more homogenous world of credit spreads (see Exhibit 1). If we think about default risk from this point of view, the resulting lower risk and longer tails can have a significant impact on correlation instruments. From the perspective of a seller of protection, lower default risk is a good thing at all levels of the capital structure, but very much so at the lower levels of subordination, given the leverage. While the longer tail negatively impacts subordinate tranches, at some point it actually benefits more senior tranches (because of diversity), depending on the length and thickness of this tail. The devil, though, is in the details, because the valuation of just a few credits can materially impact the results. With a focus on Dow Jones TRAC-X II (and similar portfolios), we conclude that the risk distribution, combined with technical factors, argues for selling protection starting with 3% to 5% attachment points for 5-year maturities and 7% attachment points for 10-year maturities. In both cases, though, we recommend combining the long credit positions with single-name delta hedges and some market spread widening protection through more senior tranches. exhibit 1
Equity Indicators – Less Risk, Longer Tails
Percent of Universe 8% Equity Market Implied Default Risk 7%
Credit Market Implied Default Risk
6% 5% 4% 3% 2% 1% 0% 0.0 0.3 0.6 0.9 1.2 1.5 1.8 2.1 2.4 2.7 3.0 3.3 3.6 3.9 4.2 4.5 4.8 Normalized Default Probability
Source: Morgan Stanley
204
THE TAILS OF THE DOW JONES TRAC-X PORTFOLIOS
From the perspective of what the credit markets went through in 2002, the long tails forecasted by today’s equity indicators are quite benign. Yet these equity-forecasted tails are still much longer than those implied by the more homogenous credit markets. According to the Merton models, the riskiest credits in the Dow Jones TRAC-X II index are Viacom, EDS, Visteon, Sun Microsystems, Sears, Delphi, Toys R Us and Computer Associates, yet only three of these names are significantly wider than the index on a spread basis. TODAY’S TAIL FAVORING 3% TO 5% ATTACHMENT POINTS
A portfolio with broader risk distribution than we initially expected would be effectively lower in correlation (or higher in diversity). This is good for “senior” tranches and bad for “subordinate” tranches (seller of protection point of view), all else being equal. To quantify the impact, we can think of this either in correlation or spread terms (see Exhibit 2). For example, the 7-10% tranche in Dow Jones TRAC-X II trades at 147 bp mid-market (18% correlation). If we then change the distribution of the spreads to match the equity-implied distribution in Exhibit 1 (keeping the average spread the same), we can get a sense of what the impact of the longer tails would be on valuation. The tranche would price at 105 bp (42 bp tighter, assuming the same correlation). Expressed in correlation terms, this is equivalent to a 6-point reduction to 12%, which is good for the seller of protection because the tranche is a short correlation position. This behavior is true for all of the 5-year tranches above a 3% attachment point. The 37% tranche should price 50 bp tighter, for example. For tranches more subordinate than this, the opposite behavior is true. For example, the longer tail implies a 5-point reduction in correlation for the 0-3%, which is bad for the seller of protection. In spread terms, the tranche should price 149 bp wider to compensate for the longer tail (at current correlation levels). The 2-5% tranche should price 36 bp wider. Longer Tails Benefit 3% Attachment Points and Higher – 5-Year Tranched Dow Jones TRAC-X NA II
exhibit 2
Tranche 0-3%
Current Market Premium (bp) Correlation 1,849 20%
With Longer Tails (Equity) Premium (bp) Correlation 1,998 15%
Difference Premium (bp) Correlation 149 -5%
2-5%
646
40%
682
33%
36
3-7%
420
5%
370
NA
(50)
-7% **
7-10%
147
18%
105
12%
(42)
-6%
10-15%
63
22%
38
14%
(25)
-8%
15-30%
15
28%
7
20%
(8)
-8%
Note: 0-3% is quoted with points up front. Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
205
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 27
Intuitively, the impacts make sense, because the longer tail in the equity indicators is based on just a few credits (i.e., the tail is not thick), meaning that 0% to 2% attachment points are the most affected. If the tail were thicker, then the risk would begin to impact tranches with 3% attachment points, or higher. In 10-year maturities, we find similar dynamics; however, the tipping point is higher in the capital structure. We therefore recommend shifting the minimum attachment higher because of the incremental risk associated with the extended maturity (see Exhibit 3). In particular, 7% attachment points appear attractive to us. exhibit 3
Tranche
0-3% 2-5% 3-7% 7-10% 10-15% 15-30%
Longer Tails Benefit 7% Attachment Points and Higher – 10-Year Tranched Dow Jones TRAC-X NA II
Current Market Premium (bp) Correlation 1,946 22%
With Longer Tails (Equity) Premium (bp) Correlation 2,099 19%
Difference Premium (bp) Correlation 153 -3%
912
30%
940
28%
28
-2%
677
30%
653
34%
(24)
4%
383
4%
251
181
17%
112
61
30%
31
(133)
**
6%
(69)
-11%
17%
(30)
-13%
Source: Morgan Stanley
LONGER TAILS WITH TIGHTER SPREADS
The longer tails, though, are only half the story if we want to use equity market indicators to make better relative value decisions. Credit risk, on average, is lower when measured by these equity market indicators. In fact, for the Dow Jones TRAC-X II index, the indicators imply that spreads should be 22 bp tighter, given the year-end relationship of the two markets. However unrealistic this may seem, if it actually happened, it would have a substantial effect on many tranches and is obviously net positive for all of the tranches (again, from the perspective of the seller of protection). We show the combined effects of spread tightening and long tails in Exhibit 4, ranging from 12 bp for the 15-30% tranche to 476 bp for the 0-3% tranche.
206
exhibit 4
Tranche 0-3%
Less Risk and Longer Tails — 5-Year Tranched DJ TRAC-X NA II
Current Market Premium (bp) Correlation 1,849 20%
With Longer Tails (Equity) Premium (bp) Correlation 1,372 (476)
Difference Premium (bp) Correlation 0-3% 1,849
2-5%
646
40%
425
(221)
2-5%
3-7%
420
5%
128
(292)
3-7%
646 420
7-10%
147
18%
39
(108)
7-10%
147
10-15%
63
22%
14
(49)
10-15%
63
15-30%
15
28%
2
(12)
15-30%
15
Source: Morgan Stanley
MODELS VS. REALITY – WHAT ARE THE TRADES TO DO?
Before we formulate ideas to implement some of the equity/credit divergence, we need to consider technical factors, as well as base-case market views. •
One danger in formulating investment strategies based on the shape of distributions is that these distributions can change rapidly (more sectors or credits become stressed). We argue that investors should keep this “tail length volatility” in mind when weighing the risks of tranched trading strategies. A significantly thickening tail, however, is not our base-case scenario.
•
We are comfortable with the notion that default risk is more dispersed than implied by spreads. This distribution, combined with current market pricing, argues for selling protection on tranches with attachment points at 3% to 5% (5year maturities). We favor this strategy with selected single-name delta hedges (to hedge thickening tails), and out-of-the-money style protection via more senior tranches or spread options (to hedge market-wide spread widening).
•
While the combination of tighter spreads and long tails is supportive of selling protection on more senior tranches, a dose of reality tells us to stick to our guns in regard to favoring buying protection, particularly for the 10-15% tranches. The models tell us that the longer tail would force this tranche to trade in the 30 bp range, and the equity-implied spread tightening would move the tranche to trade below 20 bp. We do not believe the market would allow either of these outcomes, particularly because AAA paper from competing assets (ABS and cash CDOs) would become significantly more attractive.
In a nutshell, for portfolios with characteristics similar to Dow Jones TRAC-X II, we recommend attaching long credit positions at 3% to 5% levels (for five years), given current market pricing and our views on the tail risks. We further recommend complementing this long position with selected single-name hedges and out-of-themoney protection (senior tranches or options) to hedge against significant market spread widening.
Please see additional important disclosures at the end of this report.
207
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 28
Streetwise Correlation – Taking the High Road
September 10, 2004
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Anisha Ambardar During the summer of 2003, the investment grade structured credit market experienced a significant secular shift. The market moved from the fairly active, yet disparate, world of synthetic CDOs to an explosion of liquidity in benchmark index tranches. With liquidity came transparency, the correlation buzzword, and a host of new investors (see Chapter 9 for our early thoughts). One year later, many credit investors associate the term “structured credit” directly with “correlation” or CDX tranches, even though the real market is much broader than that. This summer was not as earth-shattering, but during the quiet that marks the end of the season, a very similar market starting emerging in the high yield space. Coincidently, if we turn the clock back exactly a year, our first weekly report after Labor Day 2003 was on the topic of high yield structured credit (see Chapter 69). At that time, we noted the advances in investment grade, and questioned whether high yield was next, given the maturity of the CDO and CLO markets. A year ago, we had more questions than answers, and more issues than strategic advice. One year later, we argue that a liquid and mature high yield structured credit market is closer to a reality, for many reasons. First, there has been a fair improvement in liquidity in the underlying high yield single-name derivatives markets, along with a deeper investor base. Second, while it has taken some time, we do have one standard high yield index today. Third, trading of synthetic high yield tranches has become active, taking a cue from the very successful experience in the investment grade markets. Finally, the cash CDO markets are flush with liquidity today, more so even than during the heyday of the late 1990s. At this point, we could write a book on the importance of a well-developed synthetic structured credit market in high yield, to complement the already mature cash CDO market. As we address this topic over time, we may have enough for a book, but for now, we will focus on three points that seem most relevant to us. First, it is important to understand the characteristics of this nascent market, from pricing to the investor base. Second, the sensitivity of the benchmark high yield tranches to changes in spread and correlation is not necessarily equivalent to the standard investment grade tranches, because the tranches themselves are not necessarily similar. Finally, there is an important link between the new high yield tranches and the much more mature cash CDO market, which can aid in valuation and eventually bridge the gap.
208
THE NATURE OF THE MARKET – A FIRST GLANCE
Volumes are still thin and bid-offers have not reached the levels of investment grade tranches, but there is a reasonable amount of liquidity in the tranches of high yield CDX today, with more than a handful of dealers making markets. Market participants have picked standard tranches that are similar, in terms of attachment/detachment, to those in the cash CLO and older high yield CBO worlds. However, the 0-10% and 1015% tranches trade with points up front and 0 bp on a running basis, mainly because a zero coupon reduces the effect of the single-name default swap curve shape, which is in general harder to infer in the high yield market. exhibit 1 Tranche High Yield CDX (5 Yr)
Trading the High Yield Tranches Level 360
Delta
Base Correlation 41.50
0-10%
75/78
1
1.4
10-15%
46/49
1
2.9
42.75
15-25%
645/695
3.1
46.25
25-35%
240/275
1.7
56.25
35/45
0.3
35-100% Investment Grade CDX (5 Yr)
56.5
0-3%
38.5/39.25
2
13.3
3-7%
262/269
7.4
31.00
7-10%
104/108
3.7
35.40
10-15%
37/40
1.6
44.00
15-30%
11/12
0.5
66.50
20.50
1
Quoted with points up front and zero bp running. Quoted with points up front and 500 bp running. Source: Morgan Stanley
2
The implied base correlation curve across tranches is relatively flat (compared to investment grade), which can have some interesting relative value implications. For example, the 15-25% tranche trades just a few (3 to 4) correlation points higher than the equity tranches, but in investment grade, the 3-7% tranche is a full 10 correlation points higher. On a relative basis, this would indicate that selling 15-25% protection in high yield is more attractive than selling 3-7% protection in investment grade, based on the models. The early flows seem to be tilted toward sellers of protection in the lower tranches (010%, 10-15%), with some establishing shorts higher up by buying protection on the 15-25% tranche, which could explain the relative value we mentioned above. From what we see, hedge funds are the early players, although press reports point to participation by CLO managers who are struggling to find collateral in the loan markets.
Please see additional important disclosures at the end of this report.
209
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 28
GETTING A HANDLE ON DEFAULT RISK
With the goal of better understanding pricing and price sensitivity of the high yield tranches, an important first insight is to quantify differences in default risk. From a historical perspective, five-year high yield default rates are on average about seven times that of investment grade (assuming a A/BBB portfolio and Moody’s data), although differences could be much smaller or larger during certain points in the credit cycle. Given that expected losses are directly related to spreads in today’s standard correlation models, we often look to relative spread levels as another indicator of default risk. Today’s spread levels on the indices imply six times the expected loss for the high yield index, compared to investment grade. LOOKS SIMILAR, FEELS DIFFERENT
Why is comparing the default risk between investment grade and high yield important? For those who are accustomed to trading the investment grade CDX tranches, the standard high yield tranches may look familiar, but they will feel very different, as they were designed to be closely aligned with common tranches in the high yield and leveraged loan CDO market, rather than to mirror the risk of the investment grade tranches. From a default risk perspective, the 0-3% investment grade tranche is more like a hypothetical 0-18% high yield tranche, given today’s spread levels. But we caution that there is a lot more complexity necessary to make a direct comparison with investment grade. First, the term structure of default risk or spreads is an important factor in tranche valuation (especially junior tranches). While we have a reasonably well developed credit spread curve in investment grade, the same is not true in high yield. Thinking in terms of default rates, investment grade default risk generally increases over time, as we have shown in prior research (see Chapter 24). A similar analysis for high yield would indicate default risk that is concentrated in the early years and can actually decrease with time. Second, from a statistical perspective, the spreads of the investment grade and high yield indices act as the means of the aggregate default distributions, but the distributions can, in fact, be shaped very differently, given the disparity in the level of individual credit risk. Moreover, the shape of the aggregate default distribution is an important driver in determining pricing on tranches.
210
HIGH YIELD TRANCHE SENSITIVITY
Beyond comparisons to investment grade, the next important step is to better understand the sensitivity of the high yield tranches to changes in index spreads and correlation. We note that the range of deltas for the quoted high yield tranches is much smaller than for investment grade (see Exhibit 1). Therefore, the price sensitivity of the tranches to changes in average index spread is more clustered (see Exhibit 2). Several other points stand out to us, as well, which we characterize from the seller of protection’s perspective. exhibit 2
High Yield Tranches – Price Sensitivity
Price Change (% of Notional) 30% 20%
0-10%
10-15%
15-25%
25-35%
35-100%
0-100%
10% 0% -10% -20% -30% -40% 200
250
300
350 400 450 Average Spread
500
550
600
Source: Morgan Stanley
•
The 25-35% tranche (and all tranches more junior) are more sensitive to broad spread moves than the index itself, implying that they are levered. The 35-100% tranche has very little price sensitivity, which corresponds to the more conservative AAA-rated attachment points in the CLO market.
•
The 15-25% tranche appears to be the most negatively convex, implying that losses for a given spread move wider are more than gains for the same spread move tighter. The 0-10% tranche appears to be the most positively convex, which we would attribute to the low implied dollar price associated with the large upfront premium.
Please see additional important disclosures at the end of this report.
211
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 28
exhibit 3
High Yield Tranches – Delta Sensitivity
Delta 4.5x
0-10%
10-15%
15-25%
25-35%
4.0x
35-100%
3.5x 3.0x 2.5x 2.0x 1.5x 1.0x 0.5x 0.0x 200
250
300
350
400
450
500
550
600
Average Spread
Source: Morgan Stanley
•
Another way to observe convexity is to simply see the changes in delta (based on 10 bp spread move, see Exhibit 3). At today’s spread levels, the 15-25% tranche’s delta increases quite a bit for wider moves (i.e., negative convexity or gamma), although it does eventually flatten at the 500 bp range. The 25-35% tranche is also negatively convex, while the 0-10% tranche is positively convex across all spread ranges.
•
The 0-10% tranche is a long correlation tranche, as one would expect (i.e., spreads tighten as correlation increases, see Exhibit 4). The inflection point is the 15-25% tranche, which is upward sloping (short correlation). At current spread levels, none of the standard mezzanine tranches are correlation invariant (or non-monotonic). exhibit 4
Spread
High Yield Tranches – Correlation Sensitivity
212
10-15%
15-25%
25-35%
35-100%
5%
Source: Morgan Stanley
0-10%
10%
15%
20% 25% Correlation
30%
35%
40%
BRIDGING THE HIGH YIELD GAP
Finally, beyond the look and feel of the new high yield correlation instruments, the emergence of this market is an important secular trend within all of the high yield space, in our view. In our own research, we have been trying to bridge the valuation gap between the structured credit and cash CDO worlds for some time (see Chapters 70 and 71). Correlation and spread data from the high yield tranche market brings the correlation-based valuation of cash CDOs one step closer to a reality, although we caution that there are still material differences between the simple high yield tranches and the more structurally elaborate cash CLO/CBO securities. Cash structures have valuable triggers that are difficult to explicitly price, and the nature of the underlying collateral (maturity, spread distribution and recovery rates) can be significantly different than the more liquid high yield CDX index, particularly for older CDO transactions. Nevertheless, the gap is beginning to narrow, and we expect market participants to push it even further ahead.
Please see additional important disclosures at the end of this report.
213
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 29
A Shot at High Yield Tranches – Bottom-Up
October 1, 2004
Primary Analyst: Sivan Mahadevan Primary Analyst: Brian Arsenault Primary Analyst: Peter Polanskyj Primary Analyst: Andrew Sheets Ajit Kumar, CFA Many market participants approach the correlation space in a top down manner, given the ease in doing so with all of the liquidity in the benchmark instruments. Today, there is quite a bit of focus on trading patterns, long/short tranche ideas, and relative value in correlation skew. In the investment grade markets, this approach is certainly acceptable, given compressed spread levels and healthy corporate balance sheets, although we continue to encourage investors to consider default and spread risk in the companies that make up the tails of the indices. The structured credit opportunity in high yield, on the other hand, requires a much different approach, in our view. The mistake most made in the original high yield CBO market was to apply average market factors to very specific portfolios. As we passed through the trough of the credit cycle, many painfully discovered how different these portfolios were from the so-called market averages. The peak 10% default rate during 2001 matched the worst period during the previous recession (1991), so it should not have surprised investors, but CBO portfolios were much more high-beta in nature. As such, the collateral in most deals underperformed the “market” by large factors, although the subsequent recovery was stronger as well. High yield is very much a bottom-up market, and we feel that any structured credit investment strategy within the space needs to follow suit, however tedious that may be. In our first steps of adhering to this philosophy, we find that a large percentage of companies in the 100-name CDX universe are both free cash flow positive and have enough cash on hand to meet debt payments over the medium term. In a nutshell, the portfolio appears like more of a low-beta trade, relative to the broader market (half the portfolio is rated BB). As such, we are constructive on taking levered default risk on the index over the medium term, although we caution that there are credits in the index that concern us, fundamentally. Additionally, as we have written about recently, long/short strategies in tranches allow investors to separate default risk from spread risk, which is an intriguing concept to apply to the high yield market (see Chapter 16). From an investment strategy perspective, we identify some interesting long/short ideas that allow investors to implement relatively bullish default risk scenarios without taking too much spread risk.
214
THE BOTTOM UP VIEW – CDX VS. THE MARKET
Despite conventional wisdom, when you invest in HY CDX you are not buying the “market,” but are essentially getting exposure to a static pool of 100 names, with a higher quality profile relative to the market. CCC risk is almost equivalent, but the BB basket is 12% greater, mainly at the expense of B rated credits. This resulting decrease in portfolio default risk is significant, at least based on historical data. The 20-year average annual default rate for a mid-BB credit is 0.67%, but an order of magnitude higher 7.65% for a mid-B. Industry sector weightings are fairly consistent with the market. HY CDX – Lower Beta, or Higher Quality vs. the Market
exhibit 1
HY Index % Market Weight 39%
HY CDX % Market Weight 51%
B
47%
36%
CCC
14%
13%
Rating Distribution* BB
*Ratings exclude MCI, using S&P ratings Source: Morgan Stanley
REFINANCING DEBT – THE TERM OUT STORY
Despite cautious tones regarding our high yield overweight recommendation (see “Bumps in the Road,” September 10, 2004), Default risk should remain muted for some time. Since 2003, high yield issuers have taken advantage of investor enthusiasm to tap the market for over $250 billion in new issuance. Although the quality has decreased recently, over two-thirds of issuance since the beginning of 2003 has been used to refinance loans and bonds (see Exhibit 2). exhibit 2
2003-2004 HY Use of Proceeds, YTD
% of Proceeds 40% 2003 Total = $146 bn 2004 Total = $112 bn 30%
20%
10%
0% Capex
Stock Buyback / Dividend
Acquisition
General Corporate Purposes
LBO
Refinance Bank Debt
Refinance Bond Debt
Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
215
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 29
HOW MUCH DEFAULT RISK?
We have mentioned in previous research that high yield issuers are sitting on hoards of cash. In fact, almost 16% of the total debt for the 100 companies in HY CDX is covered entirely by cash today. In fact, since so many companies have pushed amortizations into the future, 62% of the issuers in HY CDX have debt covered by cash on the balance sheet up to the end of 2006. This stands in contrast to the issuance boom of the late 1990’s, when proceeds were primarily applied toward capex and expansion. Cash as a percentage of debt for the high yield index stood at 9% in Q1 1999, half of what it is now. In addition, 77% of the names in the HY CDX index are free cash flow positive. exhibit 3
100
Healthy Balance Sheets – % of HY CDX Companies with Debt Covered by Cash**
90 80 70 60 50 40 30 20 10 0 2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
**Debt amounts include straight debt, converts and term loans Source: Morgan Stanley
THE TAILS OF THE HY CDX INDEX
Despite the market’s strong liquidity profile, at least a few defaults are inevitable, especially within the CCC space, as the historical annual default rate for CCC issuers is in the 25% area. However, to blindly hedge all CCC credits would be a waste of carry, in our view. Issuers such as Huntsman and Crown Castle are in the process of being rewarded for operational improvement. For names such as Calpine, Six Flags, Charter, AMR, and Level 3 (to name a few), their operational challenges are reflected in yearto-date equity returns (see Exhibit 4).
216
exhibit 4
HY CDX – CCC Issuers
Years Covered 1
YTD Equity Return N/A
Name Huntsman Intl Llc
Moody’s Caa1
El Paso Corporation
Caa1 /*-
CCC+
NEG
2
6%
Caa1
CCC+
NEG
3
-37%
Calpine Corp
S&P CCC /*+
S&P Outlook POS
Meristar Hospitality Crp
B2
CCC+
NEG
3
-17%
Six Flags Inc
B3
CCC+
NEG
5
-27%
Dynegy Holdings Inc
Caa2
CCC+
NEG
6
8%
B3 /*+
CCC /*+
POS
3
32%
Chrtr Comm Hlds/Chrt Cap
Ca
CCC-
POS
1
-34%
Levi Strauss & Co
Ca
CCC
DEVELOP
2
7%
Caa1
CCC
POS
4
35%
Crown Castle Intl Corp
American Tower Corp AMR Corp Triton PCS Inc
Caa2
CCC
STABLE
NA
-44%
B3 /*-
CCC
NEG
7
-53% -54%
Level 3 Communications
Caa2
CC
DEVELOP
4
Qwest Capital Funding
Caa2
B
DEVELOP
1
-28%
Rite Aid Corp
Caa1
B-
STABLE
2
-43%
Owens-Illinois Inc
Caa1
B
NEG
1
29%
Source: Morgan Stanley
Away from the CCC credits, there are a handful of other names that warrant caution as well. For example, in Exhibit 5 we examine issuers whose amortizations are not covered by cash over the next 2 ½ years and have lagging equity performance YTD. To overstate the obvious, limited liquidity and plummeting market caps are not a recipe for success in high yield investing. exhibit 5
Name Dura Operating Corp
Early Amortizations and Lagging Equity Performance Moody’s B2
S&P B /*-
S&P Outlook NEG
Years Covered 1
YTD Equity Return -44%
Allied Waste North Amer
B2
B+
STABLE
1
-37%
Mediacom Llc/Cap Corp
B2 /*-
B+ /*-
NEG
2
-26%
Navistar International
Ba3
BB-
STABLE
1
-24%
Abitibi-Consolidated Inc
Ba2
BB
NEG
1
-21%
HCA Inc
Ba1
BBB-
STABLE
1
-10%
B2
B-
NEG
2
-3%
Collins & Aikman Prodcts Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
217
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 29
FROM FUNDAMENTALS TO STRATEGIES
One of the biggest sources of opportunities in the high yield structured credit space is that fundamental views like what we described above are not fully reflected in tranche pricing, for several reasons. First, default risk is not the only factor that drives spreads in the high yield market, even though risk-neutral (i.e., correlation) models are built upon this premise. Second, default swap credit curves are still not that well developed today for the full 100 names in the index, so the positive near-term nature of default risk implied by today’s fundamentals will not necessarily translate into curve shape and model-based valuations. exhibit 6
Levered Long Total Returns – 3 Year Holding Period
60% Sell 0-10% Protection 40%
Sell 10-15% Protection
20%
0%
-20%
-40%
-60% 0 Defaults, Spreads Tighen
5 Defaults, Spreads Flat
10 Defaults, Spreads Flat
15 Defaults, 20 Defaults, 25 Defaults, Spreads Widen Spreads Widen Spreads Widen
Source: Morgan Stanley
DEVELOPING LONG/SHORT STRATEGIES
In developing long/short investment strategies using the benchmark tranches, the first step is understanding the total return profile for levered long side of the trade. The 1015% has positive return for up to 15 defaults (versus 9 defaults for 0-10%), but it also has a larger potential negative return for high numbers of defaults (see Exhibit 6). Because of the upfront nature of payments, the 0-10% position would benefit from further rising short rates and back-ended defaults (because the larger upfront payment would earn more interest). In Exhibit 7, we illustrate the total returns of two long/short strategies under the same scenarios. The first strategy is to sell protection in the 10-15% tranche and buy an equal-notional amount of protection in the 0-10% tranche. This package does well in scenarios where HY CDX experiences 10 to 15 defaults, but despite being a long/short strategy, it is not delta hedged (which would be difficult, see below) and remains exposed to near term spread moves. Furthermore, it requires a large net upfront payment and would suffer if short rates continued to rise. Another long/short strategy is to pair the sale of protection in the 10-15% tranche with the purchase of protection in the index, on a delta-neutral basis. This package has its best performance in the 5 to 10 defaults scenarios and has much less exposure to small spread moves and positive convexity for large spread rallies. For investors looking to hedge both early default risk and market-wide spread moves, this long/short strategy may be preferable. Further, the strategy would benefit from rising short rates, because
218
there is a net positive upfront payment. The deltas, though, require close attention, because the upfront nature of payments implies that deltas can increase substantially even while the duration of the underlying index drops (see Exhibit 8). exhibit 7
Long/Short Total Returns – 3 Year Holding Period
60% Sell 10-15% Protection, Buy 0-10% Protection Sell 10-15% Protection, Buy Index Protection
40%
20%
0%
-20%
-40%
-60% 0 Defaults, Spreads Tighen
5 Defaults, Spreads Flat
10 Defaults, Spreads Flat
15 Defaults, 20 Defaults, 25 Defaults, Spreads Widen Spreads Widen Spreads Widen
Source: Morgan Stanley
exhibit 8
Watch Out for Rising Deltas
Delta 9.0
0-10% 10-15%
8.0
15-25%
7.0
25-35%
6.0
35-100%
5.0 4.0 3.0 2.0 1.0 0.0 5
4
3 Years to Maturity
2
1
Source: Morgan Stanley
CONCLUDING THOUGHTS – DEFAULT RISK VS. SPREAD RISK
While we have presented readers with quite a bit of fundamental and strategic food for thought, a key-take away is that we are fundamentally supportive of taking default risk on the 100-name HY CDX index. There are clearly many ways investors can implement this view, but we chose to focus on long/short strategies that minimize the impact of small spread moves because this represents a new and relatively interesting way to approach the high yield market.
Please see additional important disclosures at the end of this report.
219
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 30
Tranches – Navigating the Auto Storm
April 29, 2005
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Ajit Kumar, CFA Pain is certainly being felt in several corners of the credit community since the midMarch earnings guidance announcement by General Motors. Many investors were unprepared for the ensuing credit volatility, which turned out to be very high by any recent measure, although still relatively low compared to the high default rate environments of 2001 and 2002. The structured credit markets are both contributing to and reacting to the auto storm, but in ways that reflect differences in idiosyncratic exposure. The toughest part of the structured credit trade over the past several weeks has been in the equity tranches; implied correlations have fallen several points (tranches cheapening) to levels only seen several years ago when tranches were traded by appointment and bid offer spreads reflected the lack of liquidity in a then-nascent market. Therefore, long and long/short equity tranche strategies are being challenged by movement in the autos and resulting run-time risk-management decisions. exhibit 1
The Auto Storm – 5Year CDS Premiums
1,200 FMCC
GMAC
DPH
VC
1,000
800
600
400
200
0 Jan-05
Source: Morgan Stanley
220
Jan-05
Feb-05
Mar-05
Apr-05
Apr-05
On the other hand, investors higher up in the CDO capital structures are feeling much less auto-specific pain. Pricing has remained largely stable (albeit negative) in these parts of the capital structure relative to market-wide movements, pointing to the differentiation between idiosyncratic and systemic risks we have discussed in past research. exhibit 2
250 Curve Adjusted Basis (bps)
Automobiles and Components Basis Jumps
200 150 100 50 0
04/28/05
04/10/05
03/23/05
03/05/05
02/15/05
01/28/05
01/10/05
12/23/04
12/05/04
11/17/04
10/30/04
10/12/04
09/24/04
09/06/04
08/19/04
-50
Source: Morgan Stanley
EQUITY TRANCHES – THE STORM CLOUDS FORM
The most obvious impact of the auto widening and subsequent hedging activity by the holders of first-loss tranche risk is the widening of CDS premiums versus cash. Exhibit 2 illustrates this dynamic very clearly for the 5-year point. While we would expect correlation levels for equity tranches to move marginally downward (tranches getting cheaper) as any sector became distressed, the move in the autos has driven equity valuation beyond any recent lows. This is likely due to the ubiquity of auto names among the on- and off-the-run index products, in which many active market participants have positions. We note a differentiation in pricing of equity tranches based on how much exposure they have to the auto sector. The recent activity in index tranches highlights this with Dow Jones TRAC-XSM NA 2 (four autos in the portfolio) trading cheaper on an implied correlation basis than Dow Jones CDX SM NA 3 (with three autos), which, in turn, trades cheaper then DJ CDX NA 4 (also with three autos, but with American Axle replacing Delphi). This difference is all the more striking when we consider that the TRAC-X NA 2 index is 15 months shorter in maturity than CDX NA 4. The shorter maturity should drive correlation levels for the same parts of the capital structure higher, not lower as we see it in the market today.
Please see additional important disclosures at the end of this report.
221
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 30
Exhibit 3 illustrates the implied correlation levels for some of the on-the-run benchmark tranched first loss products. We use the on-the-run index, which incorporates the roll from CDX NA 3 to CDX NA 4, minimizing any time decay effect. The moves in the US investment grade indices are significant despite the fact that CDX NA 4 is marginally less risky than CDX NA 3. To give some context for the price impact of these differences, a correlation point is worth about 0.70 price points for CDX NA 4. Cheapening Equity Tranches
exhibit 3
DJ CDX IG 5Y DJ CDX IG 10Y DJ CDX HY
Tranche 0-3%
Recent (CDX 4) 15.75%
1/4/2005 (CDX 3) 18.50%
Change -2.75%
0-3%
10.25%
15.00%
-4.75%
0-10%
37.00%
34.00%
3.00%
Source: Morgan Stanley
One of the unexpected results of the auto volatility was the rapid flattening of auto curves in the short end, which we have argued has been caused by a desire for equity tranche investors (or their risk managers) to hedge what is known as jump-to-default risk, or the risk that a credit “jumps” quickly to default. One way to hedge jump-todefault risk is to buy short-dated protection versus the sale of long-dated protection in a spread-duration-neutral manner. Another is to simply “over-hedge” the autos by buying additional auto protection in the front end. In either case, a flatter curve results. MARCH – SAILING INTO THE WIND
For investors involved in the most subordinate parts of the capital structure, the recent events have tested the assumptions of many “hedged” strategies. While we experienced significant volatility in March, the performance of long/short equity tranche strategies largely depended on how single-name risks were hedged. Those using the index to hedge equity tranche positions were faced with increasingly under-hedged exposure to the autos as auto spreads widened dramatically. This proved somewhat painful (see Exhibit 3). Those who came into March with single-name hedges in place (one for each of the 125 names, based on model deltas) faced some downside but fared better as the hedges generally performed well, considering the magnitude of what occurred (see Exhibit 4).
222
exhibit 4
P/L of Long Equity Tranches vs. Delta Hedges – How You Hedged Mattered (DJ CDX NA 3)
Hedging Strategy March
5 Year
10 Year
Index Delta Hedge
-1.7%
-2.4%
Single-Name Delta Hedge
-0.5%
-1.4%
Index Delta Hedge
-4.0%
-0.0%
Single-Name Delta Hedge
-3.1%
0.8%
Single-Name Delta Hedge with Rebalance
-2.8%
1.0%
Single-Name Delta Hedge with Default Rebalance
-3.2%
0.5%
April
Source: Morgan Stanley
APRIL – SURVEYING THE DAMAGE
The real trouble came in April, though, when auto spreads paused and eventually continued their trip toward four digits (see Exhibit 1). The volatility in the autos and other high-volatility names caused a repricing of equity risk in structured credit, most notably in the 5-year maturity. As the appetite for auto risk dramatically declined (likely as investors licked their wounds from March), and correlation levels moved even lower, delta hedging did not stem losses. In Exhibit 4, we have approximated the returns of several “hedged” long/short strategies in April (assuming they were initiated at the beginning of March). The strategies are: •
Index delta hedges without any rebalancing in the period
•
Single name delta hedges without any rebalancing in the period
•
Single name delta hedges with spread risk rebalanced on April 1 (i.e., deltas are adjusted based on spread movements)
•
Single name delta hedges adding a default and spread neutralizing hedge on April 1 for the three autos
The key takeaways are fairly clear for both 5- and 10-year maturities. Index hedges performed the worst, as the dispersion we saw in the market was not captured by the simple hedge. The rebalanced single-name strategy outperformed both the static strategy and the default rebalanced strategy, in which we short 3-year risk and go long risk to the maturity of the equity position. For 10-year risk, the pattern is similar except that the single-name hedged strategies actually generated positive returns.
Please see additional important disclosures at the end of this report.
223
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 30
THE STORM IS NOT OVER
We argue that we are still in the middle of the auto event of 2005, and it is likely not a sustainable situation to have the entire auto sector trading at current market levels. So what happens to the equity tranche strategies if we exit the auto storm in either direction? If the autos continue to deteriorate (or even default), the default-neutralized strategy should outperform. In distressed scenarios, the credit curves will invert, and this strategy should outperform, as it is, in effect, a flattener. In the case of a default, this strategy is immunized by design. But the more interesting scenario is what happens if risk aversion declines in credit markets again and the auto sector recovers. In Exhibit 5, we show the hypothetical performance of the same strategies in a market where auto spreads and implied correlation levels return to those we experienced at the beginning of March (however distant that may seem). For both 5- and 10-year risk, the default-hedged strategy outperforms all the singlename hedging strategies. This serves to highlight at least one motivation for using curve trades to hedge equity exposure. While the curve trades underperform in deteriorating spread scenarios (like what we are living through now), they will eventually outperform as curves invert with extreme distress, and they will also outperform the directional single-name hedges when spreads rally significantly. The curve positions thus offer better hedges for extreme outcomes, while retaining exposure to near-term mark to market. exhibit 5
P/L of Long Equity Tranches vs. Delta Hedges – What Happens If Markets Recover?
Index Delta Hedge
10 Year -0.2%
Single-Name Delta Hedge
1.2%
2.7%
Single-Name Delta Hedge with Rebalance
1.1%
2.7%
Single-Name Delta Hedge with Default Rebalance
1.7%
3.5%
Source: Morgan Stanley
224
5 Year 1.7%
CONCLUSION – WHO CAN WEATHER THIS STORM?
As we described, the equity tranche versus default-neutralized deltas are a bit like convexity trades: they perform well in the extreme cases but not if markets do not reach the extremes. Therefore, contrary to strategies where investors are net long or short the market, big moves in either direction may be the saving grace for equity tranche investors, at least based on the four strategies we outlined. It is also important to note that this may be a market environment where size matters, as it may take time for markets to settle, and those who are in a position to put even more capital to work may find better risk and reward opportunities. Equity tranche trading strategies are levered idiosyncratic plays on credit, a point emphasized by activity over the past two months. Yet it is also important to highlight that investors who have sought safety from idiosyncratic events through higher positions in the CDO or CDO squared capital structures have so far gotten what they bargained for in this market environment – much more stable pricing relative to the market at large and the pricing models.
Please see additional important disclosures at the end of this report.
225
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 31
Correlation – Don’t Throw Opportunity Out May 13, 2005 with the Model Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Viktor Hjort Primary Analyst: Afsheen Naqvi Ajit Kumar, CFA While the storm in the auto sector has not subsided, the reaction that it triggered in the market benchmark equity tranches has now moved well beyond this idiosyncratic event to one that is much more technical in nature, garnering attention in every corner of the investment community and financial media. As researchers, our team’s time has been split among discussing activity with credit market participants, fielding questions from an incredible flood of participants from other markets, and thinking about opportunities. In this very technical market environment, there is a lot that we do not know, including the net appetite among investors who want to either enter or re-enter equity tranche trades, and the level of anxiety among those who want to get out. However obvious it may sound, the side that tips this scale will drive pricing in equity tranches and perhaps fuel market moves in single names, as well. In this chapter, we focus on three points: •
A brief description of what happened in the market standard index tranches, and also what did not happen.
•
A discussion of the warning flags that we have raised over the past several months, many of which need to be re-emphasized given the varying degree of experience among market participants.
•
Opportunities as we see them today, although we expect that technically related flows will continue to dominate pricing for the near to medium term.
WHAT HAPPENED?
226
•
The stress in the US auto sector served as a trigger point for market participants who have sold protection on 0-3% index tranches and underhedged spread and/or default risk for the auto names in question, most notably GMAC and Ford Motor Credit, but also Visteon and Delphi (see Chapter 30, where we discuss this impact in detail). Both investors and arrangers of synthetic CDO transactions can have exposure to this risk.
•
Stress in the single-name credit markets and in the convertible arbitrage community may have forced many levered investors to reallocate risk and rebalance portfolios. The result is that many investors are forced sellers of index tranche long/short strategies. The unwinding of the most popular of these is bringing to the market buyers of equity tranche protection and sellers of mezzanine tranche protection.
•
These technical flows have driven market prices of tranches. While pricing is very much a moving target and may continue to be so, implied correlation on the onthe-run CDX equity tranches had fallen from 16% to about 9% over the past weeks due to these flows and has recovered to about 11% now. Before the auto volatility, implied correlation on equity tranches was in the 19-20% range. iTRAXX equity tranche correlation moves were similar, going from 22% to 12% and recovering to 14%. Exhibit 1 summarizes the difference between actual and expected price changes (based on index performance and correlation models) for select tranches since the beginning of May. exhibit 1
0-3% (points upfront chg)
Delta-Hedged Tranche Price Changes in May (bp) 5Y US 4.0pts
Euro 3.9pts
10Y US 1.5pts
Euro -2.5pts
3-7%
(84)
(49)
133
28
7-10%
(38)
(24)
(85)
(19)
10-15%
(15)
(11)
(37)
(55)
15-30%
(2)
(1)
(14)
(6)
Note: Delta-hedged positions involve selling protection on the tranche versus buying protection on a delta-adjusted notional amount of the underlying index. Source: Morgan Stanley
WHAT DID NOT HAPPEN? LOOK BACK IN TIME
The CDO market has been active for 10 years, and the 2001-2002 turn in the credit cycle served as a very important test for the market. High yield default rates peaked at 10.4% during that time, with many high yield CBOs experiencing even higher default rates depending on the credits in their collateral pool. Synthetic CDOs (which were dominated by investment grade credits) experienced numerous fallen angel defaults, to a point where equity tranches and in some cases mezzanine tranches experienced at least a partial loss of principal. Since the 2001-2002 period, market participants have learned some important lessons, and in many ways, the existence of standardized index tranches and the long-short nature of the market is indeed a reaction to that negative credit cycle. These tools have allowed both deal arrangers and credit investors to better hedge positions and to express credit views in different ways. Yet they have also encouraged the market to grow much bigger. The auto event is idiosyncratic in nature and does not yet represent a sharp turn in the credit cycle, in our view. So why is this event such a big deal if it seems relatively benign compared to 2001-2002? The size of the structured credit market is certainly one factor, as is its liquidity and transparency, which has served as a double-edged sword for some. WARNING FLAGS HAVE BEEN RAISED
Unfortunately, many investors jump first and ask questions later, and good liquidity can sometimes make this an easy thing to do. The worst outcome of an event like this is that no lessons are learned. We are fortunately in a position to fall back on some of
Please see additional important disclosures at the end of this report.
227
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 31
our previously published research in this space to revisit the warning flags that we raised. •
Equity tranches carry a lot of idiosyncratic default risk, even if they are delta hedged with the underlying portfolio or more senior tranches. This is a key motivating factor for the trade. Dispersion, or a widening of the shape of spread distributions, has a negative impact on equity tranches.
•
In Chapter 46, we show that one credit moving 1,000 bp wider would result in a 12.6% immediate loss to the 5-year equity tranche, assuming no correlation shift. The same move would cause only a 4.1% loss in a 3-7% tranche, and a deltahedged position of equity versus mezzanine would still result in approximately a 5% loss in the position (approximately 2 to 1 delta). More than one credit moving dramatically wider or a shift in correlation could exacerbate this loss, as we have seen in the market.
•
In the same chapter, we show that an actual default could (depending on the name and assuming 40% recovery) result in a 15% loss to the equity tranche (10% if the index is delta hedged), but only a 2.4% loss to the 3-7% tranche. We also demonstrate that the impact of a default on the equity tranche of the high yield index would generally be smaller. For a default in the widest-trading high yield credit, the impact could result in a loss of less than 1% whether outright or delta hedged and one as large as 3% for the tightest credit (assuming no change in correlation).
•
In our recent experiences, we feel that investors appreciate the risk of default in these strategies, but not the risk of credits moving substantially closer to default without actually defaulting, as we highlighted above.
•
Many investors have chosen to concentrate their long/short tranche strategies in the index tranches because of the attractiveness of liquidity and the simplicity of using a product that is a market standard. However, we have cautioned in the past that doing so exposes investors to both concentrated idiosyncratic risk and technical issues that are beyond any one investor’s control (again, see Chapter 46). We have recommended that investors diversify their instruments for exactly these reasons.
•
We continue to view the auto storm as an idiosyncratic event, and pricing in the tranche markets is in many ways supporting this view. The large community of investors globally who have used synthetic CDO technology to gain exposure to credit through mezzanine and senior tranches has not been hurt significantly by price action so far. We believe it will likely take at least one if not multiple defaults for price and ratings action in more senior tranches to be significant enough to bring in sellers of such structures to the market.
WHAT CAUSES US CONCERN?
Certainly the technicals that we have highlighted above are a big source of concern, but an actual default event, even from a CCC-rated company, will bring some jitters to the market at this point, especially if it is part of the HY CDX or iTRAXX Crossover indices, where a correlation repricing could also reasonably occur. However, we note that the standard high yield index tranches are much less sensitive to a single default than investment grade, and there are far fewer long/short high yield tranche strategies implemented in the market place (see Chapter 46).
228
An investment grade default would be a much larger event and would certainly send more jitters through the market, even though some of the pain of a default is already being priced in, given where auto names and equity tranches are trading. A default in a non-auto name would be particularly troublesome. WHERE ARE THE OPPORTUNITIES?
It is difficult to make investment recommendations during a time when market technicals are dominating pricing. Equity tranches have obviously experienced the most amount of price action, and many who are involved are wondering whether current correlation levels are attractive. Equity tranches are among the simplest instruments in the structured credit market. They are also among the riskiest. Our recommendation is to push aside correlation models for the moment, and to look at equity tranche opportunities with an eye toward relative pricing and differences in the levels of portfolio jump-to-default risk. exhibit 2
% of Total 10-Year Default Risk (PV) Allocated in the First 5 Years 5/11/2005
4/29/2005
5Y Change
CDX (US) 0-3%
82%
73%
9%
3-7%
17%
23%
-6%
7-10%
17%
15%
1%
5/11/2005
4/28/2005
iTRAXX (Europe) 0-3%
68%
58%
11%
3-6%
17%
21%
-4%
6-9%
16%
17%
-1%
Source: Morgan Stanley
Market participants often hear that risk is moving into or out of tranches. We use this idea in Exhibit 2 to measure the allocation of risk for the various US and European tranches. In the exhibit, we calculate the percentage of the total expected losses for the 10-year tranches allocated to the front and back five years and how that allocation of risk has changed given recent market activity. The takeaway is that for both European and US equity tranches, risk has shifted dramatically into the front years, while for the junior mezzanine tranches, risk has moved into the back five years in both indices. PRICING IN A DEFAULT
Another way to visualize the price action on tranches is through the price moves, which we highlight in Exhibit 3. Again, the most significant price action is in the 5-year equity tranches and in the 10-year junior mezzanine tranches. It is notable that both European and US equity protection costs increased by 12-14 price points. To give a benchmark for this price move, a single default at 40% recovery would require a payment of 16 points from a holder of these tranches. So both European and US tranches have effectively priced in an additional default in the next five years with near certainty. Today, the total
Please see additional important disclosures at the end of this report.
229
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 31
upfront payments alone on CDX equity tranches effectively fund 3-4 defaults in five years (4-5 in 10 years), while the iTRAXX equity tranche upfronts fund 2-3 defaults in five years (4 in 10 years), assuming 40% recovery in default. Price Action in Tranches – CDX and iTRAXX
exhibit 3
CDX Index
5/11/2005 5 Year 10 Year 73 97
4/29/2005 5 Year 10 Year 56 83
Change 5 Year 10 Year 17 14
0-3%
58%
72%
46%
65%
12%
6%
3-7%
265
865
261
636
4
229
7-10%
65
220
67
245
(2)
(25)
5 Year 53
10 Year 76
5 Year 29
10 Year 52
5 Year 24
10 Year 24
0-3%
42%
65%
28%
55%
14%
10%
3-6%
176
585
171
450
5
135
6-9%
53
188
55
183
(2)
5
iTRAXX Index
Source: Morgan Stanley
The 10-year junior mezzanine tranches widened significantly against the backdrop of tightening spreads in most other mezzanine tranches even as the underlying indices widened. The repricing of the long-dated first mezzanine tranche seems fundamentally consistent with price action in 5-year equity: if we are pricing in an additional default in the front end, then the probability that we touch the lowest mezzanine tranche is likely much higher, particularly with longer maturity. But the technically driven strength in the 5-year 3-7% (3-6% in Europe) seems at odds with these dynamics.
230
EUROPEAN AND US EQUITY TRANCHES – DIFFERENCES ARE BROADER THAN THE ATLANTIC
While correlation markets in both the US and Europe were consistently repriced over the past few days, the technical drivers may be creating opportunity for those who believe that such technicals will eventually fade. In particular, this remarkable consistency in repricing stands in contrast to the differing risks that exist within the investment grade indices between Europe and the US. In Exhibit 4, we show a comparison of the distribution of risk within the CDX and the iTRAXX indices. We use the ratio of the average spread on the five widest names to the average spread of the portfolio to measure the magnitude of dispersion in the underlying portfolio. While there is a dramatic increase in risk for the tail of the CDX index, we do not see anything similar for iTRAXX, which leads us to question the magnitude of the iTRAXX equity repricing relative to the US. exhibit 4
US and European Tails Are Different – Spread Ratio of the 5 Widest Credits to the Index
9 8 CDX 7
Itraxx
6 5 4 3 2 1 0 Sep-03
Jan-04
May-04
Sep-04
Jan-05
May-05
Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
231
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 31
Until the recent technicals took hold, implied correlations told a similar story. In Exhibit 5 we show the difference between the implied correlation for the CDX and iTRAXX 0-3% tranches throughout 2005. We started out the period with similar tail risks in the portfolios and with iTRAXX equity tranches at a 1% correlation premium to CDX. exhibit 5
Correlation Difference – 0-3% 5-Year CDX and iTRAXX
Implied Correlation Points 0% -1% -2% -3% -4% -5% -6% -7% Jan-05
Feb-05
Mar-05
Apr-05
May-05
Source: Morgan Stanley
Going into the end of April, the CDX portfolio already represented a much more idiosyncratic play on credit and the iTRAXX tranche priced with a 7% correlation premium. With the recent technicals, that difference has come back to less than 3%, representing a significant decline from very recent levels without any reduction in the tail risk of the CDX portfolio or increase in tail risk for the iTRAXX portfolio.
232
CROSSING THE ATLANTIC WITH EQUITY TRANCHES
In a market dominated by technical flows, leaning against them is a hard strategy to recommend, but we find it tempting to at least identify a European/US equity tranche strategy that would benefit from a reversal in technicals, even though we are not calling for one. In Exhibits 6 and 7, we show a package where an investor buys protection on the CDX 0-3% tranche and sells protection on the iTRAXX 0-3% tranche, both with index delta hedges (with the caveat that index deltas have their problems). The trade will benefit from a rising European correlation relative to the US, which is consistent with the difference in portfolio composition we see in Exhibit 4. It is a bearish idiosyncratic trade in the US, combined with a neutral to bullish idiosyncratic trade in Europe, and does require an upfront payment of points. Currency hedging decisions would also have an impact, but here, we have ignored this aspect. exhibit 6
5-Year Equity Tranche Package
Position 0-3% CDX
Maturity 6/20/2010
Notional (30,000,000)
CDX
6/20/2010
390,000,000
Package
U/F (%) 58.25%
Running (bp) 500
VS 72
1.32%
40
41.04%
(20) 500
0-3% iTRAXX
6/20/2010
30,000,000
41.00%
iTRAXX
6/20/2010
(450,000,000)
0.78%
35
Package
29.25%
(25)
Package
-11.79%
(5)
53
Source: Morgan Stanley
exhibit 7
10-Year Equity Tranche Package
Position 0-3% CDX
Maturity 6/20/2010
Notional (30,000,000)
CDX
6/20/2010
390,000,000
Package
U/F (%) 73.00%
Running (bp) 500
VS 97
2.14%
65
45.21%
(345)
0-3% iTRAXX
6/20/2010
30,000,000
64.00%
500
iTRAXX
6/20/2010
(450,000,000)
1.59%
55
Package
40.15%
(325)
Package
-5.06%
20
76
Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
233
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 31
HOW DO US EQUITY TRANCHES LOOK IF WE NEUTRALIZE AUTO RISK?
As we have highlighted above, the biggest difference between US and European equity tranches is the much wider tail in the US, which is largely attributable to the two autos in question. If we neutralize this auto risk (which we describe below) in the 5-year CDX equity tranche, we find that the US equity tranche is still wider than the European equity tranche, which is one way of demonstrating that technicals in the US have gone beyond these autos, relative to Europe (see Exhibit 9). There is no perfect way to hedge the impact of one credit on an equity tranche, and as we have discussed in the past, the type of hedging employed can significantly impact an investment strategy’s performance (see Chapter 30 for more details on various hedging alternatives and their impact on performance). We examine three alternative default hedging strategies. The first is to simply buy the full notional amount of CDS protection to the tranche maturity outright. This strategy would incorporate a bearish view on the auto spreads in addition to the default hedge and would hence suffer in an auto spread rally. The second is to buy 3-year protection and sell protection to the tranche maturity such that the trade is spread neutral but still fully hedges default exposure. This strategy has less immediate directional exposure but the hedge essentially adds forward long auto spread and default risk. The third strategy combines strategies 1 and 2 in equal proportions. Depending on our decisions regarding 5- versus 10-year equity tranches, and the type of auto hedging employed, the carry and sensitivity varies. Exhibit 8 highlights one version of the strategy. In Exhibit 8, we have also attempted to compare the hedged position to an iTRAXX equity tranche with the same maturity. In this case, there is approximately 7.6% of additional upfront payment for taking risk in the hedged CDX portfolio over iTRAXX. In Exhibit 9, we show the difference in upfront equivalent payments for taking risk in similarly-hedged 0-3% tranches of CDX versus the equivalent in iTRAXX (again ignoring currency issues).
234
5-Year CDX Equity Tranche with Auto Hedges
exhibit 8
Position 0-3%
Maturity Notional 6/20/2010 30,000,000
U/F 57.25%
Running 500
DV01 (per $1000)
GMAC
6/20/2008 (20,000,000)
680
2.7
GMAC
6/20/2010 12,000,000
700
4.5
(8,000,000)
(173)
0
FMCC
6/20/2008 (20,000,000)
580
2.7
FMCC
6/20/2010 12,000,000
630
4.5
(8,000,000)
(135)
0
Package
57.25%
Equivalent w/500 Coupon
48.60%
500
Euro
41.00%
500
CDX (without FMCC/GMAC) vs iTRAXX
192
7.60%
Source: Morgan Stanley
In the 5-year tranches, there is a pick-up in the price of default risk versus iTRAXX, while the reverse is true for 10-year, where the iTRAXX equity appears to offer better compensation for taking levered default risk. While these results are not adjusted to reflect the remaining aggregate difference in risk between the CDX (without FMCC and GMAC) and iTRAXX portfolios, such an adjustment (though it would make all the CDX strategies less attractive) would not change the result that 5-year CDX strategies appear more attractive to 10-year strategies on a relative basis. exhibit 9
5 year 10 year
Equity Tranche Premium Advantage Default Hedge 6.10%
Curve Default Hedge 7.60%
Split Hedge 6.85%
-2.84%
-1.33%
-2.08%
Source: Morgan Stanley
CONCLUSION
From an opportunity set perspective, we live in a world where market technicals can continue to dominate price movements for a good time to come. Beyond highlighting what did and did not happen, along with the appropriate warning flags, we believe that it is important to emphasize what the technical pricing in the market is implying about risk going forward.
Please see additional important disclosures at the end of this report.
235
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 32
Valuing Equity Tranches in the Real World
May 20, 2005
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Ajit Kumar, CFA Given spectacular moves in equity tranche prices and the media’s obsession with it, the structured credit markets continue to occupy a spotlight in the financial investment world. The only thing that we know for certain at this point is that its days of occupying this spotlight are thankfully numbered, as market participants get on with their investment process. But for those in the investment community who are closer to this market (or are taking the time to learn), the spotlight is irrelevant, while the potential opportunities are not. We have had an enormous amount of communication with clients recently days on the topic of equity tranche price movements in the standard index tranches, having one-onone meetings and calls with perhaps 50 institutions. Hedge funds (mainly outside of credit) and insurance companies are the two client bases that are in the best position to react to the price movement, based on our recent experiences. Given a general lack of directional conviction on equity tranche prices at the moment (although equity tranche upfronts are at the tight end of an 8-point range this week), many are considering such trades from a total upside/downside perspective, indicating that there is appetite for potential mark-to-market volatility. But the key motivating factor is the idea that the standard instruments have limited downside, a bit like buying a bond at 20 cents on the dollar. The clear theme in our discussions with investors is that market pricing of standard equity tranches is implying more default risk than most can fathom, but there is a healthy debate about how to actually measure this. Ironically, this is exactly what correlation models are supposed to do, in a risk-neutral or market-implied sense, but with market pricing so technically driven (in tranches, the underlying credits and indices), we would argue that a very large risk-premium component is clouding default risk measures. In Chapter 31, we argued that current pricing in various five- and ten-year tranches implies that a fair amount of default risk is priced into the near-term. While most investors fundamentally disagree with what the pricing implies, they are still searching for better ideas about how to think about default risk over time, given that market risk premiums are a distorting factor. In this chapter, we turn the clock back several years to a more simplistic world and think about equity tranche risk relative to historical expectations of default risk (i.e., the real world), which is how many investors approached the CDO market in the early days. But to avoid ignoring everything that we have learned in the years since then, we take into consideration the cyclical nature of default risk (using simulations) and also consider portfolio sampling risk, or the risk that a given portfolio’s behavior deviates
236
from the market. Portfolio sampling risk serves to add volatility to any distribution of default risk, particularly for index-sized (125-name) portfolios. exhibit 1
5-Year Cumulative Default Rates – Moody’s Universe and a Portfolio with CDX Ratings
% 10
DJ CDX IG 5 Yr Equivalent Moody's IG Universe 5 Yr
8
6
4
2
0 1974
1980
1986
1992
1998
2004
Source: Moody’s, Morgan Stanley
HOW MUCH DEFAULT RISK? LET’S START WITH THE BASICS
There are a few characteristics of investment grade default rates that are worth noting as they are important in any analysis of real-world default risk (all based on Moody’s historical data). •
Default rates are very economically cyclical.
•
Default rates rise exponentially as ratings fall. In investment grade, there is a big difference between BBB default rates and defaults rates for AAA through A credits.
•
Ratings migration in investment grade is a net negative phenomenon, meaning that the average credit quality of a fixed portfolio declines over time, based on historical data. Among other things, this implies that default risk on a forward basis should be greater than default risk today.
•
Default rates and recovery rates are negatively correlated, meaning that when default rates increase, average recovery falls (we discuss the impact on equity tranches later).
In Exhibits 1 and 2 we show the five- and ten-year cumulative default rates for investment grade portfolios over time, based on fixed annual universes (cohorts) of credits, which takes the net negative ratings migration phenomenon into consideration. For example, the five-year default rate for the 1992 cohort was 0.08% while the fiveyear default rate for the 1997 cohort was 0.86% for the Moody’s investment grade universe. In the same exhibits, we also show the historical default rates for a portfolio with the same credit quality weightings as CDX, which is the most relevant to our subsequent analyses. Note that portfolios with CDX-like quality weightings have more volatile historical default experiences than the universe at large.
Please see additional important disclosures at the end of this report.
237
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 32
exhibit 2
10-Year Cumulative Default Rates – Moody’s Universe and Portfolio with CDX Ratings
% 10
DJ CDX IG 10 Yr Equivalent Moody's IG Universe 10 Yr
8
6
4
2
0 1979
1984
1989
1994
1999
2004
Source: Moody’s, Morgan Stanley
LET’S NOT FORGET PORTFOLIO SAMPLING RISK
A common approach early in the development of this market was to simply use historical default experience (such as in the previous section) to get a sense of the risks in CDO structures. As many learned in the 2000-2002 period, the major risk in this approach in a high default rate environment is the fact that historical data represent the experience for an entire universe of corporate credits while the risk in most structures is actually related to a relatively small set of credits. Given that the entire universe was generally much larger than any given portfolio in 2000-2002, results in many portfolios were, not surprisingly, much worse (or better) than results for the entire universe of corporate credit. As we have discussed in past research (see Chapter 63), aggregate default experience for typically sized portfolios can be significantly more volatile than what historical aggregate statistics would imply. exhibit 3
Distribution of Default Rates: Portfolio Size Matters – 5 Year Investment Grade
45% Portfolio (125 Names) 40%
Universe (1,000 Names)
35% 30% 25% 20% 15% 10% 5% 0% 0.000.25%
0.250.75%
0.751.25%
1.251.75%
1.752.25%
2.252.75%
2.753.25%
3.25- 3.75% 3.75% or More
Source: Moody’s, Morgan Stanley
In Exhibit 3, we show the results of a simulation we ran based on the last 30 years of Moody’s 5-year cumulative aggregate default experience and the rating distribution of the
238
CDX portfolio. The exhibit shows the distribution of 5-year aggregate losses for a universe of 1,000 names as well as the distribution for a randomly selected 125-name portfolio within that universe. For each trial, we randomly select a default rate for the 1,000-name universe from the actual Moody’s data from 1970 forward. We then randomly select a portfolio of 125 names and measure how many of the universe defaults are in the portfolio. Exhibit 4 shows the results from a similar analysis for 10-year aggregate losses. Consistent with our prior research in the space, we find that the 125-name portfolio, while having the same expected losses as the entire universe, has a much more fattailed distribution, with much higher probability of a zero default rate, as well as a higher probability of a default rate that is much higher than average. This portfolio sampling risk is a key risk in any structured credit transaction. exhibit 4
Distribution of Default Rates: Portfolio Size Matters – 10 Year Investment Grade
45% Portfolio (125 Names) 40%
Universe (1,000 Names)
35% 30% 25% 20% 15% 10% 5% 0% 0-0.5%
0.51.5%
1.52.5%
2.53.5%
3.54.5%
4.55.5%
5.56.5%
6.5- 7.5% or 7.5% More
Source: Moody’s, Morgan Stanley
EQUITY TRANCHES AND REAL WORLD DEFAULT RISK
The much more fat-tailed distribution of risk for 125-name portfolios helps us to examine the recent price action in equity tranches in a new light. In Exhibit 5, we show the implied pricing of equity tranches on an all upfront basis (i.e., no running premium) as of April 29, as well as the likelihood of having portfolio losses in excess of the breakeven loss rate (based on the analysis in Exhibits 3 and 4). As a point of reference we have also included the same probability for a portfolio comprised of the entire universe (which we assume to be 1,000 credits). The difference between these two is an indication of the level of sampling risk in the portfolio. It is notable that the risk/reward of the equity tranches was reasonably balanced at the time. As of April 29, market pricing told us that the breakeven loss rate for a 0-3% tranche on 5-year CDX was 1.76%, which is calculated from the upfront premium (we assume no reinvestment income from the premium, which is a conservative assumption). Based on our simulations, the probability of exceeding this breakeven loss rate was 21% (which is roughly equivalent to the sum of the probabilities to the right of the 1.25-1.75% category in Exhibit 3). For the 10-year equity tranche, the equivalent numbers are a 2.45% breakeven loss rate and a 38% probability of exceeding it.
Please see additional important disclosures at the end of this report.
239
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 32
Probability of Exceeding Breakeven Loss Rates (April 29, 2005)
exhibit 5
Tranche 0-3% 5 Year CDX
Premium (All Upfront) 58.8%
Breakeven Portfolio Loss 1.76%
Probability of Exceeding (CDX) 21%
Probability of Exceeding (Universe) 14%
81.6%
2.45%
38%
37%
0-3% 10 Year CDX
Note: All upfront premiums are not actual trading levels, but those implied by standard pricing. Source: Morgan Stanley
In Exhibit 6, we repeat the same exercise using more recent pricing (May 18, which serves as the low point of pricing) under two assumptions. For the 5-year equity tranche, the breakeven loss rate rises to 2.45%, and our simulations show that exceeding that loss rate is an event with only 13% probability (about 40% less than what we saw about two weeks prior). The 10-year results do not change much. exhibit 6
Probability of Exceeding Breakeven Loss Rates (May 18, 2005) Probability of Exceeding (CDX) 13%
Probability of Exceeding (Universe) 5%
0-3% 5 Year CDX (w/1 Default @ 0% Recovery)
36%
34%
0-3% 5 Year CDX (w/1 Default @ 40% Recovery)
21%
14%
Tranche 0-3% 5 Year CDX
0-3%10 Year CDX
Premium (All Upfront) 74.4%
87.0%
Break-even Portfolio Loss 2.45%
38%
37%
0-3% 10 Year CDX (w/1 Default @ 0% Recovery)
2.61%
63%
67%
0-3% 10 Year CDX (w/1 Default @ 40% Recovery)
49%
51%
Note: All upfront premiums are not actual trading levels, but those implied by standard pricing. Source: Morgan Stanley
The other scenario we considered is one in which the market is pricing in an immediate default in the portfolio with certainty (both at 0% and 40% recovery). The probability of exceeding the 5-year breakevens jumps to 36% and 21%, respectively, for the two recovery assumptions for immediate defaults. It is interesting to note that with one immediate default at 40% recovery, the 21% probability of exceeding the breakeven rate is equivalent to a zero-default scenario using April 29 pricing, suggesting that the market has indeed priced in one immediate default with near certainty. The impact of this single assumption on the results of the analyses is rather striking and helps to explain the price action on equity tranches in recent weeks. Without the assumption of a default, the risk profile appears to be very skewed with a very low
240
probability of losses exceeding those “expected” by the market. In the second analysis, with its assumption of one definite default, the risk/reward looks much more similar to what we saw at the end of April. EQUITY TRANCHES AND INSURANCE POLICIES
We include one more real world valuation example for equity tranches: homeowners’ insurance. Most (if not all) homeowners’ insurance policies have a deductible. The reason for the deductible is straightforward: to prevent the insurance company from being responsible for small losses that can be expected to occur with regularity. The idea is not to cover the cost of a light bulb going out but to indeed cover the loss of a roof or the loss the home to fire subject to the policy limit. exhibit 7
Index CDX CDX
Tranche 5Y 0-3%
Equity Tranches as Insurance Policies
Implied U/F Attachment Detachment 62.75% 1.9% 3.0%
Thickness 1.1%
Spread on Implied Tranche 1,342
10 Y 0-3%
74.00%
2.2%
3.0%
0.8%
1,923
iTRAXX
5Y 0-3%
52.25%
1.6%
3.0%
1.4%
1,047
iTRAXX
10 Y 0-3%
70.75%
2.1%
3.0%
0.9%
1,709
Source: Morgan Stanley
The same can be said for equity tranches. Most people would agree that we can expect some losses from the CDX portfolio but beyond a small number of defaults, the portfolio risks are much less certain. The structure of equity tranches reflects this concept with an upfront payment to reflect the high probability of some positive amount of loss on the portfolio and 500 bp of risky spread to compensate for the uncertainty of the size of the actual losses. We can think of equity tranches as insurance policies that cover losses in the portfolio up to a limit of 3% (for a 0-3% tranche) but with a deductible equal to the upfront payment received. In this framework, the 500 bp of spread on the tranche can be thought of as the premium for losses between the upfront amount and the 3% cap. While this approach is a bit conservative because it ignores any interest income on the upfront before any defaults actually occur, it is also a bit aggressive in not reflecting the riskiness of the spread. In Exhibit 7, we show the implied spread on this net position using this approach to give another indication of the price of equity risk. CONCLUSION
Despite all of the financial progress in thinking about tranche pricing from the perspective of market-implied default rates, technical factors have pushed up risk premium to a point where market-implied (i.e., risk-neutral) measures are very misleading, forcing market participants to think about the risk in much simpler terms. Given the large number of investors who are circling around this equity tranche opportunity currently, marrying valuation with real-world expectations for default risk certainly has some merit, and the results are very enlightening relative to historical expectations of default.
Please see additional important disclosures at the end of this report.
241
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 33
High Yield Equity Tranches – Isn’t It Ironic?
June 10, 2005
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Ajit Kumar, CFA The peak of structured credit volatility during May generated enough attention to raise the interest of many investors from other corners of the financial community who carefully looked at or entered the structured credit market. Auto-related stress drove market focus toward tranched risk in investment grade portfolios, which dominates outstanding volumes in the synthetic structured credit space. We find it somewhat ironic that there was a general lack of attention and relative price action within the high yield tranches during this time period, especially considering that a default event had occurred in the market standard high yield CDX index (Collins & Aikman). This disparity in market reaction is certainly a function of investment grade tranched risk being much larger in volume than high yield, but it also highlights the fundamental differences in the risk associated with the various tranches, as well as the comfort levels of those involved in the respective markets. As we have mentioned in earlier research, the high yield tranche opportunity requires an approach based on credit fundamentals (assisted by structural differences), and high yield tranche investors’ comfort with fundamentals may help to explain the market’s ability to digest a default event. After the recent market attention given to equity tranches, we find many investors today approaching equity tranche opportunities using simple statistical analyses based on historical data, much like in the early days of the CDO market. We addressed such an approach in detail in Chapter 32. In this chapter, we apply the same technique to the high yield market and attempt to shed some light on non-correlation model valuations of the standard high yield equity tranches (0-10% and 10-15%). From a historical perspective, the high yield equity tranches are much more underwater than their investment grade counterparts, but they have much lower downside and can be fairly sensitive to both the timing of defaults and positive ratings migration patterns. Though we are not discussing correlation in this chapter, these two points are justification for equity tranches trading at higher correlation than investment grade.
242
DEFAULT RATES ARE VOLATILE; SAMPLING RISK IS NOT
While it may be obvious to high yield investors, it is worth highlighting that default rates in high yield can be much more volatile than in investment grade, a phenomenon that has important implications to any statistical approach to tranche valuation. In Exhibit 1 we show the cumulative 5-year default rate for the actual Moody’s high yield universe, as well as a universe adjusted to match the ratings profile of the high yield CDX portfolio. Five-year default rates ranging from below 5% to above 35% certainly demonstrate the volatility of default rates through time. Within the investment grade space, one of our key takeaways is that there are significantly different loss distributions for 125-name portfolios (equal to investment grade CDX) compared to the entire investment grade universe, despite the expected losses being the same. Effectively, the smaller portfolios have much fatter tails than bigger ones. exhibit 1
Five-Year Cumulative Default Rates – Moody's Universe and Portfolio with CDX Ratings
40
Moody's SG Universe 5 Yr DJ CDX4 HY 5 Yr Equivalent
35 30 25 20 15 10 5
2004
2002
2000
1998
1996
1994
1992
1990
1988
1986
1984
1982
1980
1978
1976
1974
0
Source: Morgan Stanley, Moody’s
However, given higher and more volatile high yield default rates, the same portfolio sampling risk is less relevant in the high yield markets. In Exhibit 2, which shows the implied loss distribution for portfolios similar to the high yield CDX index, we find that the difference between the distribution of losses for the high yield CDX sized portfolio and one containing the entire universe is much more muted relative to investment grade. The implication for tranched credit markets is that extreme events on either tail are less likely for high yield portfolios than for investment grade, regardless of portfolio size. Offsetting this is the fact that a much more broad range of loss rates can occur in high yield with reasonably high probabilities (versus investment grade).
Please see additional important disclosures at the end of this report.
243
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 33
exhibit 2
Distribution of High Yield Expected Default Losses
30%
DJ CDX4 HY 5 Yr (100 Names)
Universe (1000 Names)
25% 20% 15% 10% 5%
>3 0%
27 -3 0%
24 -2 7%
21 -2 4%
18 -2 1%
15 -1 8%
12 -1 5%
912 %
69%
36%
0%
Source: Morgan Stanley, Moody’s
It is worth noting that the distribution for high yield aggregate losses in Exhibit 2 has two humps (bi-modal), which highlights the risk in using average default rates in analyzing high yield portfolios. Our interpretation of this is that periods of relatively high or low aggregate losses are much more common than periods of “average” experience. HIGH YIELD TRANCHES AND REAL WORLD DEFAULT RISK
Using the historical analysis of default rates we introduced above, we can now develop some insight into how resilient (or not) the standard high yield CDX equity tranches are to losses. In Exhibit 3 we show both break-even loss rates and the probability of exceeding those loss rates for the 0-10% and 10-15% tranches (from a protection seller’s perspective).
244
exhibit 3
Probability of Exceeding Break-Even Loss Rates
Upfront Premium
Break-even Portfolio Loss
Probability of Exceeding: CDX
Probability of Exceeding: Universe
4/29/2005
79.0%
7.9%
88.4%
98.5%
5/17/2005
87.5%
8.8%
84.5%
96.2%
6/7/2005
81.3%
8.1%
88.4%
98.0%
4/29/2005
48.5%
12.4%
58.6%
55.4%
5/17/2005
70.5%
13.5%
55.3%
52.0%
6/7/2005
59.3%
13.0%
58.6%
53.7%
0-10% Tranche
10-15% Tranche
Source: Morgan Stanley
The probability of exceeding the break-even loss rates for the 0-10% tranche is quite high, above 98%, if we assume the full universe of high yield credits, which clearly demonstrates how in-the-money (or underwater) this tranche feels. However, if we reduce the universe to 100 names, this probability drops about 10% to the 88% zipcode, though it remains very high. For the 10-15% tranche, probabilities of exceeding the break-even loss rates are still in the 50% range despite the 10 percent of subordination. COMPARING INVESTMENT GRADE AND HIGH YIELD
Correlation model users will usually claim that equity tranches feel like in-the-money options on default, meaning that they are underwater and expected to lose money. In Chapter 32, we calculated the probability of exceeding the break-even loss rate on the 0-3% 5-year investment grade equity tranche to be approximately 20%, which compares to approximately 60% and 90% for the high yield equity tranches, respectively. Despite being much “thicker” than their investment grade brethren and offering all upfront cash flows, the high yield tranches represent (at least from a historical perspective) much riskier plays on credit. Exhibit 4 illustrates a comparison of historical portfolio expected losses to tranche breakeven losses for several CDX tranches. For high yield tranches, expected losses are likely to exceed break-even thresholds, whereas they are not likely to do so for investment grade 5-year and 10year equity tranches.
Please see additional important disclosures at the end of this report.
245
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 33
exhibit 4
Expected Losses vs. Tranche Breakeven
16 14
Ratings based expected losses
12
Equity tranche breakeven losses
10 8 6 4 2 0 IG CDX 4 - 5 yr
IG CDX 4 - 10 yr
HY CDX 4 - 5 yr 0-10%
HY CDX 4 - 5 yr 10-15%
Source: Morgan Stanley
THE TIMING OF DEFAULTS CAN BE IMPORTANT
We make the conservative assumption above that no reinvestment income is earned on the large upfront payments in the high yield tranches. Because the high yield tranches offer much larger upfront payments and have no running premium, the impact of this assumption is much more important here than in investment grade and should be taken into consideration when comparing equity tranche opportunities. If we instead assume that all defaults happen in the back half of the trade (last 2 ½ years), then reinvestment income (at the risk free rate) would result in equity tranche loss probabilities falling about 4% for 0-10% and about 2% for 10-15%. The best-case scenario would be defaults happening near maturity of the tranche, in which case reinvestment income would lower the probability of losing money on the equity tranche by 15% to 75% for the 0-10%, but still only 2% for the 10-15%. POSITIVE RATINGS MIGRATION
Default rates generally increase exponentially as we go down in credit quality within high yield. Therefore, any ratings-based analysis of expected losses can be considerably influenced by ratings changes within the portfolio. The analysis above assumes that the rating migration of high yield CDX follows historical patterns. Significant positive deviations from these patterns can introduce some large decreases in the likelihood of exceeding the break-even loss rates. For example, if we assume that all credits rated below B- (16 out of 100 in CDX) migrate upward to single B, then history tells us that the probability of exceeding the breakeven threshold on the equity tranche would fall nearly 20% to 69% for the 0-10% tranche, and about 10% to 47% for the 10-15% tranche (see Exhibit 5). Other positive ratings migration scenarios are interesting, as well (again, see Exhibit 5).
246
exhibit 5
Reduction in Probability Due to Changes in Portfolio Quality
25% 0-10% Tranche 10-15% Tranche
20%
15%
10%
5%
0% All CCC upgraded to B
20% B upgraded to BB
20% BB upgraded to BBB
Source: Morgan Stanley
CONCLUSION
As we have stated numerous times in the past, the tranched high yield opportunities are very different than those in investment grade. There are differences in underlying universe risk, the differing effect of portfolio size and structural differences in the standard traded products. As our analysis has demonstrated, high yield equity tranches are much more underwater than their investment grade counterparts, but they have much less downside, putting asymmetry on the side of the seller of protection, a bit like buying low dollar price bonds. This phenomenon alone is attractive to many types of investors, providing one of the justifications for higher implied correlation. If we use historical default experience as a guide, high yield equity tranches are bullish plays on both the timing of defaults and on positive ratings migration, especially at the CCC level.
Please see additional important disclosures at the end of this report.
247
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 34
90% Fundamentals – The Other Half Is Correlation
June 17, 2005
Primary Analyst: Sivan Mahadevan Primary Analyst: Brian Arsenault Primary Analyst: Peter Polanskyj Primary Analyst: Andrew Sheets Ajit Kumar, CFA As some level of normalcy returns to the structured credit market following the spring cleaning event triggered by stress in the auto sector, three interesting themes jump out at us. First, a lot more investors are looking at structured opportunities today than before, given that the attention the market garnered raised curiosity levels and, in many cases, resulted in putting some points on the board in an otherwise difficult trading environment. Second, there is significant interest in the market today for noncorrelation model approaches to valuing structured credit opportunities. Third, and perhaps most important, we continue to view opportunities in the high yield corner of structured credit as very interesting fundamental plays on credit quality and default risk. We have argued in the past that risk-neutral (correlation) models must be interpreted much differently in the high yield space, given both the extreme volatility and cyclicality of default rates (see Chapters 16 and 29). In today’s environment, fundamental views on the tail of the HY CDX index, along with views on what could be tomorrow’s tail credits, can have meaningful implications on tranche investment strategies. While we acknowledge that such opportunities exist in investment grade, the sheer volatility of default rates in high yield can lead to a large range of possible outcomes. In this chapter, we team up with our colleague, high yield strategist Brian Arsenault, and provide an update on our views on the standardized high yield CDX tranches from a fundamental perspective. In structured credit parlance, we are most concerned with significant shifts in credit risk into and out of the tails of the portfolio. Based on this analysis, we find most appealing tranche investment strategies where we have the flexibility to write-off the riskiest names and focus on hedging the tighter names that show potential to deteriorate significantly. We feel the net exposures of these strategies are both interesting and unique ways to position high yield investments. FOCUSING ON THE RIGHT TAIL
In Exhibit 2, we show the 26 credits that make up the right (risky) tail of the high yield CDX 4 indices, based on a combination of spreads and ratings. If we ignore fundamentals for the moment, market pricing implies that these 26 credits should experience a cumulative 45% default rate (about 12 credits) over five years. Based on Moody’s data, historical ratings implied default rates are remarkably similar (43%), but, given the small number of names and high level of default risk, the actual outcome could very easily be considerably higher or lower (see Exhibit 1). Adding in some fundamental opinion, our credit analysts (in aggregate) expect that about one-third of these 26 credits are more likely than not to become average high yield credits over the next three years. While these statements may seem contradictory, they actually
248
highlight the risks in hedging already distressed names within a larger tranched strategy. In some respects, the asymmetry in credit is flipped; if a stressed name is hedged, further deterioration without a default may have limited upside, while a dramatic rally in that credit may have a significantly negative impact on returns. exhibit 1
Distribution of 5-Year Triple-C Default Rates
18% 16% 14% 12% 10% 8% 6% 4% 2% 0% 0%
10%
20%
30%
40%
50%
60%
70%
80%
90% 100%
5 Year Cumulative Default Rate
Source: Morgan Stanley, Moody’s
To put these expectations in context with tranched markets, in Chapter 33 we calculated the break-even default rate for the 0-10% tranche to be approximately 12 defaults over five years. If we consider our analysts’ constructive view on a handful of the tail credits, the 12 expected defaults may fall to about eight, but the remaining 74 names in the index are still risky, with market implied default rates pointing to about 11 additional defaults. This leaves the 0-10% tranche severely underwater. However the volatility around these expectations is also generally very high and the leverage in the tranches only exacerbates it. Therefore, we find that investors need to have very strong conviction with regard to credit views and a tolerance for losing about 20% of notional (the maximum downside of the 0-10% tranche) to sell protection in the most subordinate portion of the capital structure (0-10%). For those without iron stomachs, it is likely best to seek subordination and move higher up in the capital structure in search of relative value. These tranches can represent more levered exposures to the index but have a less direct exposure to the extreme tail of the credit distribution. WRITING OFF THE RIGHT TAIL
Given the nature of the riskiest credits in the portfolio, we are tempted to simply write off the tail by assuming that it will generate losses similar to those implied by the market (about 12 defaults out of 26 credits). Under this assumption, we find the strategies of selling protection on the 10-15% or 15-25% tranches appealing, given that they can easily withstand these defaults and have additional margin for defaults generated from the rest of the portfolio. Furthermore, the more constructive aggregate views our analysts have on the tail credits would, if realized, benefit protection sellers in 10-15% and 15-25%.
Please see additional important disclosures at the end of this report.
249
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 34
Right Tail Credits – HY CDX 4
exhibit 2
Name Calpine Corp Level 3 Communications
Moody's/ S&P Caa3/CCC
Years Covered 2
% of Debt Due <5 Years 53%
YTD Equity Return -15.0
Spread 1,743
Ca/CC
3
56%
-28.0
1,460
Triton PCS Inc
Ca/CCC-
4
32%
-33.9
1,437
Charter Comm. Holdings
Ca/CCC-
1
53%
-52.2
1,425
Caa2/CCC
6
39%
22.2
1,292
AMR Corp
Caa1/CCC+
3
64%
-57.0
1,245
Tembec Industries Inc
B3/B
4
29%
-47.0
1,078
Amkor Technology Inc
B3/B-
2
67%
-25.2
946
Caa1/CCC
4
49%
-18.3
805
Delphi Corp
B3/B-
4
47%
-43.5
716
Intelsat Ltd
Caa1/B
3
5%
N/A
658
Dura Operating Corp
Six Flags Inc
Visteon Corp
B3/B-
5
73%
-27.4
619
Levi Strauss & Co
Caa3/B-
7
5%
31.7
584
Toys R Us
571
Ba2/BB
10
14%
29.2
AK Steel Corp
B1/B+
4
45%
-52.7
568
Rite Aid Corp
Caa1/B-
0
59%
16.9
520 500
B3/B
1
35%
5.1
Meristar Hospitality Crp
B2/CCC
2
69%
-1.0
492
Qwest Capital Funding
Caa2/B
2
53%
-17.1
474
Smurfit-Stone Container
B2/B
3
8%
-40.2
440
Goodyear Tire & Rubber
B3/B-
5
51%
-2.7
436
Mediacom LLC/Cap Corp
Polyone Corp
B3/B+
0
65%
-26.6
435
Caa2 /CCC+
5
33%
5.6
426
Fairfax Financial Hldgs
Ba3/BB
7
11%
-3.8
422
Bombardier Inc
Ba2/BB
2
81%
20.1
394
Caa1/B+
1
36%
-11.9
370
Dynegy Holdings Inc
Allied Waste North Amer.
Note: “Spread” is the median bond spread (bp). “Year covered” measures how many years the company can fund bond and loan amortizations with current cash and equivalents. Source: Morgan Stanley
However, the major risk in these two tranches remains the scenarios where tighter trading names in the current universe either experience enough deterioration to default or to drive negative credit migration (i.e., they become tomorrow’s tail credits). This type of migration would by no means be an outlier event, given the cyclical nature of default risk (the entire portfolio is still high yield), and so we focus our efforts on identifying the most likely candidates in today’s portfolio to be the stressed credits of tomorrow. FUNDAMENTALLY ANALYZING THE REST OF THE INDEX
In Exhibit 3, we identify 16 credits from the core part of the HY CDX index that are candidates for negative shifts in credit quality, based on a variety of criteria. The
250
objective criteria included issuers having three years or less in debt coverage, 5-year default swap premiums tighter than 400 bp, and either negative or declining positive free cash flow. Additionally, after conversations with our sector analysts, we added the following three credits, which warrant extra caution based on credit analyst opinions: Saks, a retailer with weak fundamentals, facing an SEC investigation; Nova, a highly cyclical company that would face difficulty if the economy slows within five years; and Texas Genco, a private entity with limited disclosure and a small cash position. The Left Tail – Tight Trading Names Worth Hedging
exhibit 3
Name Arvinmeritor Inc. Case New Holland Inc.
Moody's Ba1
S&P BB
Years Covered 2
% of Debt Due <5Years 62%
YTD Equity Return (%) -11.1
5yr CDS 302
Ba3
BB-
1
66%
-3.2
CSC Holdings, Inc.
B1
BB-
3
45%
8.0
N/A 195
Nortel Networks Corp.
B3
B-
3
91%
-25.6
315
Saks Incorporated
B1
B+
3
39%
27.5
385
Unisys Corp.
Ba1
BB+
3
100%
-31.9
280
Abitibi-Consolidated Inc.
Ba3
BB-
1
42%
-27.7
372
Allegheny Energy Supply
Ba3
B
0
38%
27.5
187
Dillard’s Inc.
B2
BB
3
37%
-5.2
292
Dole Foods Co.
B2
B+
3
70%
N/A
280
Caa1
CCC+
3
30%
6.5
365
B1
B
1
31%
-2.0
330
Starwood Hotels Resorts
Ba1
BB+
1
49%
-2.1
125
MGM Mirage Inc.
Ba2
BB
1
67%
17.0
190
Nova Chemicals Corp.
Ba2
BB+
1
58%
-30.5
230
B1
B
0
18%
N/A
200
El Paso Corporation Felcor Lodging LP
Texas Genco LLC Source: Morgan Stanley
TRANCHE TRADING STRATEGIES
Based on our fundamental work from above, we recommend combining the sale of protection in the 10-15% and 15-25% tranches with delta hedges in the credits identified in Exhibit 3. The devil, though, is always in the details. Given the differences in risk between the tranches, we examine two hedging strategies: the first is a focused strategy incorporating a small number of names, which suits the fairly concentrated risk in the 10-15% tranche; the second is a more diverse hedging strategy incorporating all of the names in Exhibit 3, which better matches the risk profile of the 15-25% tranche with its higher attachment and “thicker” exposure to the index. DELTA HEDGING AND INCREMENTAL DEFAULT RISK
In Exhibit 4, we summarize the tranche positions, the structure of the two hedging strategies and the implied carry positions. For the concentrated hedge, we hedge the full potential notional exposure to the first six names in Exhibit 3, based on the aggregate opinion of our analysts. For the broader hedge, we select a 5% hedge ratio
Please see additional important disclosures at the end of this report.
251
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 34
for each of the names. The 5% hedge ratio is based on our analysis of the model-driven deltas for single names, as shown in Exhibit 5. It is worth noting that the amount of single name hedging required for any given tranche differs for different names, depending on their spread levels. As with investment grade tranches (see Chapter 44), models imply more hedging notional for wide names in junior tranches, while implying more notional for tight names in senior tranches. The 5% hedge ratio is a rough approximation, and was motivated by pragmatic considerations (i.e., ease of implementation). Tranche Trading Strategies
exhibit 4
A
B
Sell Protection on DJ CDX 4 HY 10-15% Tranche
Notional ($m) 5.0
Buy Protection on 5 tight names
6.0 ($ 1m each)
Net carry
-1.3% (plus upfront $3m)
Equivalent carry after converting upfront to running
17.4%
Sell Protection on DJ CDX 4 HY 15-25% Tranche
10.0
Buy Protection on 16 tight names
8.0 ($ 0.5m each)
Net carry
4.6%
Source: Morgan Stanley
exhibit 5
What Is the Right Single Name Delta?
Hedge Notional Required
10-15% Tranche
8%
15-25% Tranche 7% 6% 5% 4% 3% 2% 1% 0% 200
250
300 Underlying CDS level
350
400
Source: Morgan Stanley
We calculated carry for the 10-15% tranche two ways, to appropriately reflect the impact of the upfront payment received by the protection seller. One way of calculating the carry is to consider interest income on the upfront (we used one year Libor), which was more than offset by the hedging cost and resulted in a negative carry of 1.3%. Alternatively, if we convert the upfront 60 points to running equivalent, the carry would be 17.4%, including the cost of delta hedging. The latter is more reflective of the true carry, in our view. Since 15-25% does not have an upfront payment, we calculated
252
the carry as the difference between the running tranche premium and cost of hedging the 16 names, which amounted to 4.6%. In Exhibit 6, we show one year total return of long 10-15% and 15-25% positions paired with their respective hedges, as discussed earlier, under several scenarios. For the tail of the index, the expected (market-implied) loss over one year is approximately 3 defaults. The 10-15% tranche plus deltas strategy outperforms the 15-25% trade in the scenarios we picked, due to its significantly larger carry. On an absolute basis, a scenario where a hedged name widens significantly (1,000 bp) resulted in the best performance for the 10-15% tranche. While the 10-15% trade looks much better from a return perspective, it is important to consider that it also carries higher default risk because of its lower attachment point. Performance can be negative in these strategies under two important scenarios. The first is where near-term defaults (from unhedged tight trading names) exceed about 6-7 credits over one year. Second, these two strategies also have negative exposure to market-wide spread moves, despite the partial delta hedges. Such a scenario, for example if spreads widen by 200 bp in a parallel fashion, results in significantly worse performance for the two tranches (see Exhibit 6 for details). exhibit 6
One Year Performance of the Two Strategies under Different Scenarios
1 yr total return (% of tranche notional) 35%
Long 10-15% 31.9%
Long 15-25%
30% 25% 20% 15% 10%
12.0% 9.8% 7.2%
6.0% 3.3%
5% 0%
-0.1%
-5% -10%
-9.7%
-15% 3 defaults + spreads roll down the curve
-10.7% -10.7%
3 defaults + 5 3 defaults + 5 3 defaults + all 10 tight names default + unhedged hedged tight spreads widen spreads by 200 bps tight names names widen unchanged widen
Source: Morgan Stanley
CONCLUSION
As we have discussed many times over the past year, the high yield structured credit opportunity is one where fundamental views on significant shifts (in either direction) of credit quality can have meaningful implications on investment strategies. When balanced with correlation trading techniques, the net credit exposure is both an interesting and unique way to approach the high yield market. From a long credit perspective, we find opportunities with 10-15% and 15-25% index tranches the most appealing, when balanced with appropriate fundamentally driven single-name hedges from the core part of the index.
Please see additional important disclosures at the end of this report.
253
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 35
Super Senior Serenity
July 8, 2005
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj For a variety of reasons, super senior tranches get the least amount of attention among all of the standardized and bespoke structured credit instruments. While the amount of notional exposure in the top-most part of the capital structure is large, the community of investors who are active in it is not, limited traditionally to select insurers and reinsurers, with some recent entrants from other corners of the credit markets. Ironically, the efficiency of funding available through super senior tranches is the reason why investment grade synthetic CDOs displaced their cash CDO counterparts nearly four years ago, which eventually led to the development of the structured credit market as we see it today. Super senior funding efficiency is a key part of the growth of synthetic structured finance CDOs as well. Simply put, those serene super senior tranches might deserve more attention than they actually receive. Pricing super senior tranches is a bit tricky. The risk-neutral correlation models tell us that the probability of penetrating them is so low that they don’t deserve much, if any, premium (partly because of the fixed recovery assumptions that are standard today), but market participants will always charge a certain amount of premium to take the risk. In a market where the other pieces of the puzzle are very liquid and transparent, super senior tranches can be thought of as the plug in a tranche versus the index arbitrage analysis, but such an analysis is by no means simple. We focus on super senior tranches now partly because the structured credit volatility that peaked in May has left such tranches at wider spread levels, despite a strong rally in equity tranches since then. We find current market pricing interesting, given what the market is saying about risk allocation to this tranche. However, while the probability of penetrating super senior tranches due to defaults remains low, other risk factors can influence pricing, including market technicals and the pricing of comparable alternatives.
254
WHY IS SUPER SENIOR CHEAPER TODAY?
During the structured credit volatility in the spring, the relative strength of mezzanine tranches stood in stark contrast to the much-watched sell-off in equity tranches, which became a painful experience for many. While there was not much focus on it, super senior tranches (30-100% on the 5- and 10-year CDX indices) widened significantly as well, from 3 bp to 9 bp for 5-year maturities, and 11 bp to 18 bp for 10-year maturities. The equity tranche sell-off brought in many buyers of credit risk, who were willing to take on a large amount of idiosyncratic risk as well, to a point where equity tranches retraced a good portion of the sell-off. Equity tranches remain the more common place for bullish plays on credit. The same cannot be said for super senior tranches, with current mid-market spreads much closer to the May wide levels than the April monthend levels. A levered exposure to super senior tranches is also a bullish play on credit, but quite different from equity tranches, as the risk of significant super senior spread widening is equivalent to increases in systemic credit risk. There is little idiosyncratic credit risk in these tranches. exhibit 1
Before, During and After Spring Volatility – Equity Tranches Come Back, Super Seniors Do Not (bp)
CDX Index
4/29/2005 5 Yr 10 Yr 58 84
5/18/2005 5 Yr 10 Yr 76 102
7/7/2005 5 Yr 10 Yr 63 87
0-3%
44 pts
66 pts
61.5 pts
74 pts
49.25 pts
67 pts
3-7%
242
643
273
1040
158
737
7-10%
62
249
70
205
49
178
10-15%
29
101
33
89
28
83
15-30%
11
43
20
46
18
41
30-100%
3
11
9
18
7
15
Source: Morgan Stanley
From a funded perspective, any credit instrument that trades north of Libor and has a small amount of default risk is indeed interesting, especially compared to the many corporate bonds and high-rated structured finance instruments that actually trade through Libor. However, the real attractiveness of super senior tranches is the ability to apply leverage (in funded or unfunded form), knowing that it takes 30% losses on an investment grade pool (i.e., 50% default rate at 40% recovery) to trigger the first dollar of loss. The historical average 5-year investment grade default rate is closer to 1% (2% for BBB-rated credits). From this perspective, the biggest risk in investment grade super senior tranches is mark-to-market risk driven by changes in the risk premium that the market will require.
Please see additional important disclosures at the end of this report.
255
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 35
PRICING IS DIFFICULT
While the pricing of super senior tranches often causes correlation models to stumble, the issues that models have valuing these tranches are more reasonable once we understand the key risks in these tranches. Risks in super senior tranches are driven by tail events in terms of both recoveries and default rates for a given portfolio. While today’s standard models assume that recovery rates and default rates are independent, historical experience tells us that in periods of high default rates, lower than average recoveries are the norm. This is consistent with the Merton model view of the world: when prices for corporate assets are broadly in a slump, more equity holders will exercise the put options provided by bondholders. So holders of super senior risk own the risk of extremely high default rates in a given portfolio, which could be exacerbated by low recoveries in such an extreme environment (think of a repeat of the Great Depression as a comparable). It is not all that surprising that our models have trouble tying with observed market prices. We not only assume recovery to be fixed but also that it is independent of default rates in today’s standard models – meanwhile risk in super senior tranches is concentrated precisely in the scenarios where one, if not both, of these assumptions break down. The trouble is that no matter how unlikely the events that would harm super senior tranches are, they are still possible, particularly for a specific 125-name portfolio. TRANCHE ARBITRAGE – THE FUNDING GAP
Given that investors have easy access to the full capital structure of standardized index tranches, there is an arbitrage relationship between the tranches and the underlying index than can aid in the pricing of super senior tranches, although it does have limitations. The ‘funding gap,’ or spread between selling protection on the index and buying protection on all of the tranches (such that notional amounts are equivalent), is generally negative and can serve as a tool to highlight relative richness or cheapness in tranches, in aggregate. In Exhibit 2, we show the historical relationship of the net spread for the strategy of selling protection on the CDX index and buying protection on the various tranches at prevailing market prices (reflecting the appropriate bid/offer and any upfront payments) along with the absolute level of the index during the same period. The strategy has consistently negative carry as it is positively impacted by default events (as we discuss below). With the volatility in the structured credit markets in May, putting this package on would have become more attractive had it not been for the repricing of super senior tranches, as the index widened by 20 basis points (which amounts to roughly 0.79 price points on the index) versus 0.66 index points of price decline in the weighted combination of equity and mezzanine tranches. This net positive effect was more than offset by a combination of widening super senior spreads and widening bid offer spreads making the funding gap more negative.
256
exhibit 2
Tranches vs. the Index Arbitrage – The Funding Gap
CDX Spread (bp)
Funding Gap (bp) 0
(2)
90 Funding Gap
80
CDX Spread
70
(4)
60 50
(6) 40 (8)
30 20
(10) 10 (12) Sep-04
Nov-04
Jan-05
Mar-05
Apr-05
0 Jun-05
Source: Morgan Stanley
BUT NOT REALLY ARBITRAGE
This arbitrage analysis, however, is only a first order approximation as any relationships between the tranches and the index have differing exposures to jump-todefault events in the underlying portfolio. To illustrate the mismatch in default exposure, consider the performance of the hedged position (long CDX protection in tranches versus short protection in the index) after a single default. The actual principal payments caused by the default are indeed hedged; however, the premium payments are not. The premium payments on the index portion of the package would be reduced by the spread on the defaulting name (about 1/125th or 0.8% of the total running spread on the portfolio), while the spread payments on the tranched portion of the package would be reduced by almost 7.5 times as much (approximately 6.0% of the total spread, based on recent market pricing). Assuming a default occurs in the next year, the value of the first default can be as much as 8 bp in price terms. In this analysis, we reflect any actual upfront payments on the index and the 0-3% tranche (and any funding costs), hence, so long as the strike prices on the index components are fixed at the same amount, the actual trading level of the name that defaults has little impact on the economics in a default. This mismatch in the premium cashflow risk serves to benefit the hedged position in the case of default and hence provides a justification for the consistently net negative carry on the index-versus-tranches package. The nature of the single-name risk should have caused the net negative carry to be greater in May, given the higher implied probability of a single default as priced by the market. The market did indeed reflect the higher level of idiosyncratic risk, which is demonstrated by a more negative funding gap.
Please see additional important disclosures at the end of this report.
257
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 35
exhibit 3
Tranches vs. CDX Arbitrage – The Impact of a Cheaper Super Senior
bp
Funding Gap
2
Funding Gap (Fixed Super Senior)
1 0 (1) (2) (3) (4) (5) (6) (7) (8) (9) 01-Apr-05
Super Senior Cheapens with Market Volatility 28-Apr-05
25-May-05
21-Jun-05
Source: Morgan Stanley
One of the key drivers of the net carry declining to current levels has been the repricing of the super senior portion of the capital structure. To illustrate the point, we show the actual net carry position for the 5-year CDX index versus tranche package under two scenarios (see Exhibit 3). The first assumes that super senior tranche pricing is fixed (at April 29 levels), while the second uses observed market levels. The takeaway is that without the widening of super senior tranches, the index versus tranche package would have become a positive carry trade, which would not have made sense in a market environment characterized by higher event risk. Given the new entrants in equity tranches and the continued technical flows in mezzanine tranches, the only mechanism to bring the relationship of the index and the tranches into balance was the repricing on super senior risk. WHAT ARE WE MISSING? TECHNICAL FACTORS AND PREPAYMENTS
As the risk of a loss due to defaults is very small in an investment grade super senior tranche, key issues for any investor taking levered exposure to the tranche are mark-tomarket risks. As correlation models do not necessarily play an important role in influencing the pricing of super seniors, market technical factors, which can be hard to measure, are a key risk. Any levered long super senior strategy must take this risk into consideration, although anecdotally, there are floor-like levels where traditional risktakers in the insurance/reinsurance industries have sold super senior protection in the past.
258
A less well understood detail is the reduction in premium (based on a smaller notional amount) that sellers of super senior protection receive after each portfolio default. While this phenomenon is not necessarily true in every super senior opportunity, for the standard index tranches, when there is a default, the top of the capital structure is written down by recoveries, while the bottom (equity tranche) is written down by losses. For example, following the Collins & Aikman bankruptcy (where recovery was approximately 45%), a high yield CDX 35-100% tranche on $650 million notional (assuming $1 billion portfolio, $10 million per name for 100 names) would be reduced in notional size by $4.5 million, thereby reducing net premium payments (albeit by only 0.7% of the original premium) to the protection seller. To sum it all up, we find that super senior tranches perhaps receive significantly less attention in the marketplace than they deserve, given their importance in determining relative value relationships and supporting synthetic CDO issuance. We find current levels on market standard super senior tranches interesting, based on a lack of a correction following the spring structured credit events, although we caution that key risks are more technical than fundamental in nature.
Please see additional important disclosures at the end of this report.
259
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 36
The Good Stuff Is on the Top Shelf
July 29, 2005
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Super senior tranches continue to be a main focus in synthetic structured credit land in the quiet of this summer, which stands in contrast to the little attention these top shelf instruments have historically received, despite their importance to the development of the market. Simply put, what has piqued interest in many investors’ minds has been wider spread levels (in the aftermath of the spring structured credit volatility) and the ease with which leverage can be applied, given the remote nature of default risk (see Chapter 35). There are some points within the super senior world that are important to understand for anyone thinking about reaching up to the top shelf. Only a small community of investors, including some monolines and select reinsurers, has the capacity to take enough notional exposure for the attractive return on capital to be scalable. As such, what this small investor base is doing (or not doing) has an impact on super senior spreads. Our surveillance suggests some activity among them, but with many focused on longer maturities and waiting for perhaps wider spreads. Other factors that could be at play include competition from alternative market segments including cash CDOs (negative basis trades, see below) and super seniors from the burgeoning world of synthetic ABS CDOs. Funded super senior investors have been non-existent until recently, but wider spreads and the ability to apply leverage is a key motivator.
© 2005 Morgan Stanley
260
With the case well established now for super seniors being intriguing from a default risk perspective, sorting through the top shelf for an appropriate opportunity can be quite challenging. Opportunities cross the funded and unfunded divide and differ in many dimensions, including underlying collateral, term, attachment and detachment points – and various bells and whistles. SUPER SENIORS IN THE CASH CLO WORLD
For those not involved in the cash CLO markets, super senior risk in this world is worth understanding, as it has some important technical implications for the synthetic CDO market. Many of the buyers of AAA CLO tranches have historically demanded wide spreads, given their own costs of funding and business models. But other investors have entered the market over the past two years or so, including balance sheet providers who are able to take large amounts of funded super senior credit risk either because they have cheap sources of funding and/or because they can hedge it in the unfunded space through a super senior credit default swap. This latter hedging exercise generally leaves them with a handful of basis points for making the effort, along with the counterparty credit risk of the protection seller. Such negative basis trades have played an important role in tightening AAA CLO paper (new issues are in the L+25 bp range today) and bridging the gap between funded and unfunded opportunities. They have also created competing products for the unfunded super senior takers in the synthetic CDO world. SORTING THROUGH THE TOP SHELF – RELATIVE RISKS
While market pricing is available for a variety of super senior opportunities in the market place, there is no standard method that can be used to compare them, given that the vast majority of the pricing is pure risk premium and correlation models do not play an important role in valuation. There may be many schools of thought on assessing risks in super senior tranches, but we focus on two in this chapter. The first is akin to thinking about super senior risk as catastrophe risk, in that the event that triggers a loss may have never actually occurred historically, at least based on Moody’s historical default rate data. In this approach, we estimate how much one has to stress the worst-case historical default experience to trigger a loss in a super senior tranche. The second approach is market based, where we stress the markets’ expected losses (instead of worst case) based on current spread levels to get a sense of how much it takes to trigger a loss. Since the best information about the distribution of portfolio risk is the market for tranched risk itself, using a spread based worst-case scenario as a benchmark doesn’t really have any teeth in a relative value context. In both cases, we attempt to measure how far out of the money these very out of the money tranches actually are. In Exhibit 1 we show both the historically driven worst-case loss rates and the current market implied expected loss rates based on the CDX portfolio.
Please see additional important disclosures at the end of this report.
261
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 36
exhibit 1
Worst-Case and Market-Implied Cumulative Losses
Index 5 Yr CDX 4
Moody’s “Worst Case” 3.8%
Market Implied 2.4%
7 Yr CDX 4
5.3%
3.8%
10 Yr CDX 4
6.2%
6.0%
25.4%
14.5%
5 Yr HY CDX 4 Source: Morgan Stanley, Moody’s
STRESSING WORST-CASE LOSSES
If we dig into the annals of Moody’s default rate histories (1970 to today), we cannot find an event that would trigger even some of the most aggressive super senior tranches in the marketplace. In fact, if we create 125-name portfolios calibrated to the ratings profile of the CDX indices to address the sampling risk inherent in specific portfolios, we find that worst-case loss rates range from 3.8% to 6.2% for 5- to 10-year terms (assuming 40% recovery). These loss rates roughly represent the CDX 125-name portfolio equivalent of the worst historical default rates experienced for investment grade credit in the last 30 years. Anecdotally, the worst-case loss rate among the actual transactions (investment grade corporate collateral) we have witnessed in the marketplace is about 4% over a near 5-year period. In Exhibit 2, we show the severity of losses necessary to trigger various attachment points as a multiple of our estimate of worst-case losses in Exhibit 1. For example a 10% attachment point for 5-years would require 2.6x the worst-case default experience to trigger a loss, while a 30% attachment would require a 7.8x worst case. Such relationships may help in determining relative value. For example, a 30% attachment point for 10-years is roughly equivalent to a 20% attachment point for 5-years, from a worst-case loss perspective. Because we are using the tail of the loss distribution as a benchmark in this analysis, these multiples partially reflect the higher expected volatility for longer term realized default rates for the portfolio (see Chapter 18). exhibit 2
Attachment 40.00%
5 Yr CDX 10.4x
7 Yr CDX 7.6x
10 Yr CDX 6.4x
5 Yr HY CDX 1.6x
30.00%
7.8x
5.7x
4.8x
1.2x
20.00%
5.2x
3.8x
3.2x
0.8x
10.00%
2.6x
1.9x
1.6x
0.4x
Source: Morgan Stanley
262
Stressed Worst-Case Losses – Super Senior Trigger Points
Taking the analysis one step further, we examine the current spreads for various senior tranches against this “Worst-Case Multiple” in Exhibit 3. For purposes of the exhibit, the super senior spreads are either observed super senior spreads or are generated by bootstrapping the observed spreads with the traded subordinate tranches. For example, the 15-100% spread is based on the weighted average spread of the 30-100% and the 15-30% tranches. This analysis gives us one way to compare super senior like tranches across maturities and attachment points, as well as for differing underlying collateral types, relative to historical worst-case default scenarios. Comparing Worst-Case Loss Multiples and Super Senior Spreads
exhibit 3
Super Senior Premium (X-100%) 8
Maturity 5
Attach 7%
Worst-Case Multiple 1.8x
CDX 4
5
10%
2.6x
7
CDX 4
5
15%
3.9x
6
CDX 4
5
30%
7.8x
5
CDX 4
7
7%
1.3x
14
CDX 4
7
10%
1.9x
13
CDX 4
7
15%
2.8x
11
CDX 4
7
30%
5.7x
10
CDX 4
10
10%
1.6x
19
CDX 4
10
15%
2.4x
17
CDX 4
10
30%
4.8x
13
HY CDX 4
5
25%
1.0x
29
HY CDX 4
5
35%
1.4x
23
Portfolio CDX 4
Source: Morgan Stanley
STRESSING EXPECTED LOSSES
Worst-case scenarios, however, are not the only drivers of risk, even though they play an important role in super senior tranches. For those who are inclined to think of the market as a better predictor of defaults than ratings-based approaches, we also perform a similar analysis but instead of comparing the tranche attachments to ratings-based scenarios, we compare the attachments to the expected losses priced into the portfolio spreads. In Exhibit 5, we show the results of the analysis, which is broadly consistent with the historical analysis. The missing piece of the puzzle for this market-based approach is the fact that the uncertainty around these expected losses is likely different depending on collateral, as well as maturity. To illustrate the point, consider that 10year 30-100% risk is similar to 5-year 20-100% risk, based on the historical approach (see Exhibit 1). However, using the market-based approach (see Exhibit 4), 10-year 30100% risk feels more like 7-year 20-100% risk, while 5-year 20-100% risk seems much more out of the money.
Please see additional important disclosures at the end of this report.
263
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 36
Stressed Market Implied Losses – Super Senior Trigger Points
exhibit 4
Attachment 40.00%
5-year 16.3x
7 Year 10.5x
10-year 6.6x
5-year HY 2.8x
30.00%
12.2x
7.8x
5.0x
2.1x
20.00%
8.2x
5.2x
3.3x
1.4x
10.00%
4.1x
2.6x
1.7x
0.7x
Source: Morgan Stanley
Comparing Market Implied Loss Multiples and Super Senior Spreads
exhibit 5
Super Senior Premium (X-100%) 8
Maturity 5
Attach 7%
Expected Loss Multiple 2.9x
CDX 4
5
10%
4.1x
7
CDX 4
5
15%
6.1x
6
CDX 4
5
30%
12.2x
5
CDX 4
7
7%
1.8x
14
CDX 4
7
10%
2.6x
13
CDX 4
7
15%
3.9x
11
CDX 4
7
30%
7.8x
10
CDX 4
10
10%
1.7x
19
CDX 4
10
15%
2.5x
17
CDX 4
10
30%
5.0x
13
HY CDX 4
5
25%
10.2x
29
HY CDX 4
5
35%
14.3x
23
Portfolio CDX 4
Source: Morgan Stanley
EXPLAINING SUPER SENIOR PREMIUMS
While the above described metrics are fairly simplistic, we find that they do play a role in determining the risk of any given super senior opportunity. For example, a simple regression shows that they explain roughly 20% to 30% of the price for a set of standard super senior tranches (see Exhibit 6). This may seem a bit low but is not surprising to us, given the highly technical nature of the market. As we can see from Exhibit 6, maturity is a key driver of relative value (based on worst-case multiples), with 5-year super seniors appearing richer, as is underlying collateral (high yield appearing cheaper). We argue that this phenomenon is consistent with the volatility of losses (10-year more volatile than 5year, high yield more volatile than investment grade). This result is remarkably consistent with the idea that these tranches are out of the money options, for which volatility of the underlying asset is a key driver in valuation.
264
exhibit 6
Super Senior Premiums vs. Worst Case Loss Multiples
bp 25 HY CDX 4 35%
20
CDX 4 10yr 10% CDX 4 10yr 15%
15
CDX 4 7yr 7% CDX 4 10yr 30%
10
CDX 4 7yr 10% CDX 4 7yr 15%
5
0 0.0x
CDX 4 7yr 30%
CDX 4 5yr 7% CDX 4 5yr 10% CDX 4 5yr 15%
1.0x
2.0x
3.0x 4.0x 5.0x 6.0x Worst Case Loss Multiple
CDX 4 5yr 30%
7.0x
8.0x
9.0x
Source: Morgan Stanley
CONCLUSION – DEFAULT RISK VS. MARKET TECHNICALS
As the risk of default in market standard super senior tranches is remote (based both on historical and risk-neutral measures), we struggle to find a good fundamental methodology to assess super senior opportunities. Market technical factors are important drivers of pricing, including the residual effects of structured credit volatility in the spring and activity among the traditional monoline and reinsurance communities. The largest risks in super senior tranches are potential mark-to-market losses, which can be influenced by market technicals and have been experienced by traditional investors in the past. From a relative value perspective, we are comforted with the notion that observed pricing in standard tranches and the two measures of default risk remoteness discussed in this chapter show a reasonably stable relationship.
Please see additional important disclosures at the end of this report.
265
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 37
Super Senior Frenzy
February 3, 2006
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Pinar Onur When super senior tranches widened out in the aftermath of correlation repricing last spring, we noted that this important but traditionally sleepy corner of the market had some interesting opportunities for investors who both had the capacity and could deal with the technicals (see Chapters 35 and 36). The increase of super senior activity by both traditional reinsurers and hedge funds, combined with the popularity of the levered super senior trade, has pushed super senior spreads to optically very tight levels, moving last year’s serenity theme more toward a frenzied sentiment. While much of the activity is in bespoke tranches, we can see the impact in the index super senior market, given dealer hedging activity and the like (see Exhibit 1). In a relative value context, the market has come a full 180 degrees from the summer of last year when the equity tranche rally pushed risk into super senior tranches. Are super senior tranches too tight? How does one assess value in this part of the capital structure? We have in the past shown how many multiples of historical loss rates are required to touch such tranches, which in some sense impacts pricing, especially when comparing tight trading investment grade super senior to wider trading high yield. But quantifying risk moving within various parts of the capital structure is easier said than done, as any funding gap (or arbitrage) relationship requires a view on the timing of defaults. Index Tranche Super Senior Levels (bp)
exhibit 1
Term 5 Yr
Tranche 30-100%
7/7/2005 (Mid) 7
2/3/2006 (Mid) 2
IG CDX 5
7 Yr
30-100%
NA
3.75
IG CDX 5
10 Yr
30-100%
15
4.75
HY CDX 5
5 Yr
35-100%
34
16
Index IG CDX 5
Source: Morgan Stanley
TRANCHE INDEX ARBITRAGE
While there is no way to perfectly hedge positions in index tranches versus the index or single name credits, the earliest and most common way to hedge similar risks has been to build complete capital structures from portfolios and tranches, like filling in both the asset and liabilities of a new issue CDO. A key difference between index tranches today and residual pieces in early deals (or equity in cash CDOs) is the fixed nature of the premium cashflows in equity index tranches. If an equity tranche receives a residual cashflow then a given deal or capital structure is hedged both from a principal and coupon perspective (i.e., whatever one
266
loses on the portfolio one gains on the tranche protection). The same cannot be said for the investment grade index tranches today, which have fixed coupon payments. As such, the strategy of selling index protection and buying protection on all of the index tranches, while being flat principal payments, would remain exposed to changes in premium cashflows caused by the timing and magnitude of default events. This makes pure arbitrage difficult in the highly liquid index tranche market. IG SUPER SENIORS – ACTUALS VS. IMPLIED
In Exhibit 2, we show the implied super senior spread generated for the CDX portfolio by selling protection in the index and buying tranche protection for the 0-30% portion. This position leaves the investor exposed to the same principal loss exposure as a standard super senior tranche. This implied premium is generally negative today (and only modestly positive in the past) because the coupon cashflows on the package are much riskier than those of a standard super senior tranche. To understand this risk, consider the impact of a single default on this implied spread. The inbound cashflows will be reduced by 45 bp (the index deal spread) on the defaulting credit notional amount while outbound cashflows will be reduced by 500 bp (the equity tranche premium) on the loss generated from the default, resulting in a net positive change in carry. The net premium could rise to be flat or be even positive depending on the recovery (we approximate a 40% recovery default to be worth about 2 bp of spread). Since the market prices in some number of defaults and losses greater than zero in any investment grade portfolio, a negative carry on the index vs. 0-30% trade can be warranted. As defaults rise, the index vs. 0-30% trade may perform better than super senior, despite the lower or even negative carry. exhibit 2
IG CDX - Replicated Super Senior (bp)
10
IG CDX 5 yr
8
IG CDX 7 yr
6
IG CDX 10 yr
4 2 0 -2 -4 -6 -8 -10 Jan-05
Mar-05
Jun-05
Aug-05
Nov-05
Jan-06
Source: Morgan Stanley
HY SUPER SENIORS – ACTUALS ARE STABLE RELATIVE TO IMPLIEDS
The relationship in the high yield world is actually the opposite of investment grade, both because there is no running coupon on the equity tranches and higher likelihood of defaults. Instead of trading at levels less than the quoted super senior tranches, implied super senior positions (sell protection on the index, buy 0-35% protection) generate spreads that are generally higher than market quotes (See Exhibit 3). This replicating position is also exposed to changing premium cashflows driven by defaults.
Please see additional important disclosures at the end of this report.
267
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 37
However, in this case defaults reduce the net spread. The relationship between actual and implied super senior spreads has compressed in the last several months (see Exhibit 3). At today’s levels we approximate it would take 2 defaults (assuming it occurs halfway to maturity) to reduce the present value of the replicating super senior premium to be equal to the quoted, which compares to 14 defaults in October. For investors looking at opportunities to get long implied super seniors, we recommend waiting for this historically volatile relationship to widen out. exhibit 3
HY CDX – Actual vs. Replicated Super Senior
150
Implied SS (Libor +Spread) Actual
100
50
0
-50 Aug-04
Nov-04
Jan-05
Apr-05
Jul-05
Oct-05
Jan-06
Source: Morgan Stanley
THE PRICE OF THE RISKY COUPON
By observing the actual traded levels on super senior tranches and the implied levels for the replicating trades, we can arrive at an estimate of what the market is charging for the uncertainty related solely to the coupon cashflows. Based on anecdotal pricing for super senior tranches over the last 9 months we summarize the average difference between the quoted and the implied premium for investment grade CDX super senior tranches for varying maturities in Exhibit 4. As we would expect, the price charged for the running spread risk increases with maturity. exhibit 4
Actual Minus Replicated Super Senior Premium (Avg. for IG CDX, bp)
5/1/2005-9/1/2005
5Y 4.4
7Y 6.1
10Y 14.6
9/1/2005-Present
4.3
7.3
14.3
Source: Morgan Stanley
As we stated above, actual super seniors trade wider to replicated super seniors because the market expects a certain number of defaults over any time period. Implementing a replicated super senior strategy requires a sufficient number of defaults within a specific amount of time. In investment grade this does not strike us as the best strategy in this credit environment. In high yield, the volatility in the relationship could create opportunity in the future as it has in the past.
268
NAVIGATING TODAY’S MARKET
In a rallying market, the relationship between actual and implied super senior tranches on investment grade CDX has been reasonably stable. Whether absolute levels are too tight is really a technical question, reminding us very much of the senior mezzanine situation in the past. The growth of levered super senior activity in the market could, in fact, reflect a structural change in the general market for super senior risk. Historically, the low stated spread and the large notional sizes of super senior tranches have constrained the risk-taking to a limited number of players. Levered super senior techniques make this type of risk accessible to both the levered community and investors who in the past have purchased senior mezzanine paper. As much as we are tempted to suggest buying protection on investment grade super senior tranches as a cheap hedge, all indications in the current market environment are for these tranches to stay tight, given the popularity of the levered super senior trade and other trends in structured credit. But for those who consider other tranches (like 710% IG 5) as a macro hedge, index super senior tranches at these levels could be interesting, partly for technical reasons. In a bad macro environment, it will be easy for contrarian investors to sell protection on 7-10% (on the wire), but levered super senior is a more complicated trade, which could keep super senior spreads wider on a relative basis to things like 7-10%. Given the notional size, a few basis points of technically driven widening could have a meaningful impact.
Please see additional important disclosures at the end of this report.
269
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 38
Six Months vs. Six Credits
October 21, 2005
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Much market attention has been given to the significant shift in risk profile of the standardized investment grade CDX indices (series 4 to 5, see “Credit Roll – Hold the Tail,” September 9, 2005). Ironically, the credits that effectively transitioned out of investment grade and into the new HY CDX index (series 5) did little to alter the risk profile of the latter. The optical pricing differences between HY CDX 4 and 5 tranches is influenced both by the credits and the 6 month extension in maturity, which, when calibrated for high yield spreads, is a period when the market “expects” anywhere from 1% to 2% losses (2-4 defaults at 50% recovery), depending on which 6-month period is considered. The same cannot be said for investment grade. Clearly there can be a lot of volatility around the default experience of any high yield portfolio, with the cyclical nature of defaults being the most important contributor. Whether one agrees or disagrees with today’s market-implied default rates over the next five years has everything to do with forming a view on where we are in this credit cycle. The large difference in pricing between similar mezzanine and senior tranches in CDX 4 and 5 is all about this cycle and time decay phenomena, and a view on them can help in formulating trade ideas. In contrast, equivalent mezzanine and senior tranches in investment grade CDX 4 vs. 5 are much more similar in pricing optics, showing how time decay is a less important factor in this corner of the market. TIME DECAY, HIGH YIELD AND OPTION PRICING
Theta, representing time decay, is one of the important Greek letters used in option pricing theory, and it generally refers to the value an option gains or loses due to the passage of time. While there is some distance between fundamental high yield analysis and option pricing theory, the concept of time decay is a fairly important variable in the pricing of high yield CDX tranches. Today’s HY CDX 5 is implying nearly 18% of cumulative losses over the five year period (see the “Market Implied” curve in Exhibit 1). As a side note, this cumulative default loss curve gets its slightly non-linear shape from the upward sloping curve that we estimate for the HY CDX 5, based on the curve shape of 26 credits (only a 5-year maturity trades for the index). The extent to which realized defaults differ from those currently priced in the market will affect both the high yield index and the index tranches. For tranche investors, it is very important to have a view on how likely this deviation is.
270
exhibit 1
Market Pricing Versus Reality – HY CDX 5
Cumulative Loss (%) 25 1996 Cohort 1998 Cohort
20
Market Implied 15
10
5
0 1
2
3 Years
4
5
Source: Morgan Stanley, Moody’s
HIGH YIELD DEFAULT RATES ARE VOLATILE
History can help us in forming this view. As we have discussed in the past, historical five-year default rates (calibrated to fit the ratings distribution of HY CDX) are fairly volatile (see Exhibit 2). If the Moody’s high yield universe were itself tranched in same way as CDX, it is clear that depending on timing either none of the standard mezzanine and senior tranches or up to 3 of these tranches could experience principal losses, depending on average recovery. Equally important to HY tranched investors is the timing of defaults, or alternatively, where we are in the business cycle. Going back to Exhibit 1, we illustrate the cumulative losses over 5-years for the Moody’s universe (again fitted to the ratings profile of HY CDX), starting in both 1996 and 1998. The five-year period starting in 1996 ended just as the US expansionary period ended, while the five-year period starting in 1998 went through the last US recessionary period. The performance relative to what is implied by today’s markets is clear. The 1996 cohort outperformed dramatically, while the 1998 cohort underperformed dramatically. exhibit 2
Five-Year Cumulative Default Rates – Moody's Universe and Portfolio with CDX Ratings
40
Moody's SG Universe 5 Yr DJ CDX4 HY 5 Yr Equivalent
35 30 25 20 15 10 5
2004
2002
2000
1998
1996
1994
1992
1990
1988
1986
1984
1982
1980
1978
1976
1974
0
Note: Default data is backward looking. Source: Morgan Stanley, Moody’s.
Please see additional important disclosures at the end of this report.
271
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 38
ILLUSTRATING TIME DECAY
While investors in most investment grade tranches are really trading out-of-the-money risk and hence mostly risk premium, investors in HY tranches face more default risk across the capital structure. Thus, while examining expected performance of investment grade tranche positions based on roll down and spread scenarios, assuming no defaults may seem reasonable, doing so for high yield tranches is much less so given that that both market and historical data indicate numerous default events per year. What that means for performance in high yield tranches is that underlying spreads matter but the timing of defaults may matter more. exhibit 3
Flat Curve (No Roll Down)
HY CDX 5 Tranche Performance 1 Year Forward Index 3.4%
0-10% 6.5%
10-15% 12.0%
15-25% 10.1%
25-35% 3.2%
Roll Down
5.0%
9.3%
18.1%
14.1%
4.3%
Roll with 1.8% Losses (Average)
5.0%
5.3%
13.8%
7.7%
4.1%
Roll with 1.8% Losses (Worst)
6.5%
10.1%
21.1%
11.9%
5.1%
Source: Morgan Stanley
To illustrate the timing issues, we show in Exhibit 3 the implied return on high yield tranches 1 year forward, in a scenario with no defaults and no shifts in implied correlation. Additionally, we illustrate scenarios where the portfolio actually experiences losses approximately equal to those priced into the portfolio in the first year. While all these scenarios generate positive returns, high yield tranches would likely underperform under a variety of scenarios including no realization of roll down, losses exceeding 1.8%, HY CDX spreads moving wider than the forwards and adverse moves in implied correlation. This illustrates the multiplicative impact of time decay as it relates to the high yield tranches. As with options, the shorter the time to maturity, the lower the price on the option – and the greater the volatility on the underlying asset, the greater this drop in price is. In high yield tranches, the very volatile nature of default rates makes the time decay much more valuable on an absolute and a relative basis than in investment grade tranches. This impact is in addition to the benefit associated with any tightening in high yield spreads for 4 year instruments versus 5 year. Investors who are long tranches are effectively betting on both the expected default events over the short term and the level of forward spreads. PUTTING IT ALL TOGETHER – INDEX TRANCHE PRICING
We have frequently been asked by investors to explain the relative pricing in on-the-run vs. off-the-run HY CDX tranches. In Exhibit 4 we show recent pricing in the HY CDX 4 and 5 tranches, which optically seems large. The 47 bp difference in the implied spread on the two indices can be attributed to three factors: the 6 months of maturity, 6 names and the basis between the traded index price and the intrinsic price of the underlying names for both portfolios. We qualify these estimates with the fact that the limited liquidity in single name CDS makes such an exercise inherently imprecise.
272
exhibit 4
Index
HY CDX Mezzanine and Senior Tranche Optics (Mid, bp) HY CDX 4 367
HY CDX 5 414
Difference 47
15-25%
368
518
147
25-35%
59
107
48
35-100%
23
37
14
Source: Morgan Stanley
Given these limitations, we approximate the impact of each of these factors on the spread differential of 47 bp between the two benchmark HY CDX portfolios. We attribute approximately 25 bp (of the 47 bp) to net increase in risk of the 6 credits, 13 bp due to the differing maturity, and 9 bp due to the difference in the basis between the indices and the underlying portfolios. If we multiply out the tranche deltas by the index premium differences, we can roughly justify the mid-market premium differences of the tranches, but the attribution of this difference is important. The CDX 4 tranches have approximately 0.5% less subordination than their names suggest, given the Collins & Aikman default earlier this year. So what are the trades to do? We have been watching the difference in pricing for the mezzanine and senior tranches in high yield indices since both series 3 and 4 started trading. The “series basis” for these tranches is often more volatile than we feel is justified given the marginally different portfolios and maturity differences in the indices, creating opportunities in various parts of the capital structure at different times. We note that the 15-25% tranche on HY CDX 4 and 5 trade at about a 150 bp basis mid-market today, rallying from about 200 bp recently. The 25-35% tranche premium difference between both indices is in the high 40 bp range. We recommend watching both of these relationships closely for good entry points.
Please see additional important disclosures at the end of this report.
273
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 39
The Big Picture Behind Relative Value
November 18, 2005
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Relative value in the structured credit world has emerged as an important business opportunity for many investors in the credit markets, thanks in part to the huge boost to liquidity and transparency from the now 2 ½ year old index tranche market. Yet, the most commonly used frameworks for today’s markets actually come from different market environments, influenced by yesterday’s performance more so than today’s. While many investors confidently talk the talk of tranches, we do find a lack of big picture focus in this corner of the market. What are you trying to get done? Looking for relative value is the wrong answer, as very few trades are truly market neutral. Most tranche ideas express some kind of credit view – be it long, short, long/short on a forward basis, or idiosyncratic risk vs. systemic risk, and the list goes on. exhibit 1
Time to Change Your Ways: Index Tranche Performance – 2004 vs. 2005
Investment Grade 5 Year 0-3%
2005 (Q1-Q3) -6.6%
3-7%
6.7%
3.8%
7-10%
4.3%
1.8%
10-15%
1.7%
0.5%
15-30%
-0.2%
-0.5%
Investment Grade 10 Year 0-3%
10.7%
-11.7%
3-7%
12.3%
-9.8%
7-10%
3.3%
4.1%
10-15%
3.2%
3.2%
15-30%
0.8%
1.4%
High Yield 5 Year 0-10%
1.8%
-15.7%
10-15%
12.1%
-17.8%
15-25%
14.2%
8.3%
25-35%
4.3%
6.2%
*HY Tranches 2004 Performance is for 4Q 2004 only. Source: Morgan Stanley
274
2004* 11.4%
We have devoted a decent amount of effort to finding opportunities in tranches over the past nearly three years, having adjusted our approaches over time given both the growing depth in the market and a reasonable amount of credit volatility. We define tranche relative value as finding attractively priced opportunities to express our credit views, i.e., the big picture. However, while a big picture focus is important, categorizing trades based on the legs of the trades is still an important way to think about them, so we describe our thoughts below and focus on a few interesting relationships in the rest of this chapter. CLASSIFYING TRANCHE IDEAS
Tranche capital structure trades – There has been a lot of focus in the market on capital structure trades involving tranches within the same portfolio or more commonly the same index. The credit risk in these trades does not involve credit differences, instead the risk is more related to the ultimate level of defaults and shifts in risk premium (i.e., spreads). One theme this year has been the shift in risk from equity to super senior and back, with the rest of the capital structure acting more stable. We dig into this in more detail below. Tranche term structure trades – These trades involve getting long or short credit risk on a forward basis, which in many cases can be more of a macro trade based on credit cycle lengths. We find less attention given to ideas like this, and as such significant shifts in risk over time are possible (see “Structured Credit Ideas – Sell Protection 10Y 3-7% vs. Buy Protection 5Y 0-7%,” October 18, 2005, for our favored idea). Portfolio composition trades – These trades involve identifying pricing discrepancies between tranches with similar risk profiles (and similar ratings), but differing underlying credits. These “incremental” credit differences can range from just a handful of credits (say 3-7% in CDX 5 vs. 4) to something more substantial (5-year 710% in IG CDX vs. 5 year 25-35% in HY CDX). Opportunities in this category are not always long/short ideas but can provide metrics for making the case for an absolute long or short investment. We consider this space to be among the most fertile in the structured credit markets, as it opens the door for bespoke activity using index tranches or other bespoke tranches as reference. Implied correlation can be an important metric in this context, useful as a guide for comparing different portfolios and maturities within the same underlying asset class but correlation needs to be supplemented with common sense as well. CAPITAL STRUCTURE – EQUITY VS. SUPER SENIOR
One of the most important capital structure dynamics this year has been the interplay of equity and super senior tranches, as everything in the middle has been much more stable. Since super senior tranches do not trade as frequently as the rest of the index tranche capital structure, we find it helpful to do a little index/tranche math to determine how risk (spread) shifts in the super senior. In Exhibit 2 we show implied super senior premium (before adjustment) that make selling protection on the index and buying protection on all the tranches a flat carry trade. Given larger running premium on equity tranches versus the index strike spread and the likelihood of at least some defaults in investment grade over time, this fair value should generally be negative.
Please see additional important disclosures at the end of this report.
275
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 39
exhibit 2
Implied Super Senior Pricing Follows Actual Pricing
Implied 30-100% Spread (bp)
5yr DJ CDX
10yr DJ CDX
8 CDX 3-4 Roll
CDX 4-5 Roll
6 4 2 0 -2 -4 -6 -8 -10 Oct-04
Jan-05
Mar-05
Jun-05
Aug-05
Nov-05
Source: Morgan Stanley
Interestingly, the patterns experienced through 2005 closely mimic the actual (rather massive) re-pricing that occurred in super senior space both with equity tranches in the spring, and in the summer when super seniors rallied back. Currently, implied super senior pricing is rather tight, consistent with current market levels. This fact, combined with the perpetual tightness of mezzanine and senior spreads seems to indicate cheapness in equity tranches, although we would argue that more cheapness is required to bring in the dollar-price buyers of risk that participated in May. We advocate watching closely, but waiting for entry. PORTFOLIO COMPOSITION – MIND THE IG/HY GAP
As we have argued recently, ratings are an important driver of senior tranche spreads and relationships (see Chapter 55). The relationship between similarly rated tranches with investment grade and high yield collateral has been fairly persistent over the last several years with high yield tranches trading consistently wide to similar investment grade tranches. In Exhibit 3, we show the basis between on-the-run investment grade and high yield senior tranches through several rolls in both indices. In some respects the positive basis is justified given the more volatile and cyclical nature of high yield default rates and the relative level of defaults given typical attachment points.
276
exhibit 3
HY and IG Senior Tranche Basis
CDX 3-4 Roll
CDX 4-5 Roll
200 175 150
Rolldown
125
Rolldown
100 75 50 25 0 -25 Aug-04
Nov-04
Feb-05
May-05
Aug-05
Nov-05
HY 5yr 25-35% vs. IG 10yr 10-15%
HY 5yr 25-35% vs. IG 7yr 7-10%
HY 5yr 25-35% vs. IG 5yr 7-10%
Linear (IG 5yr 7-10%)
Source: Morgan Stanley
Note that the high yield senior tranches have rallied quite dramatically versus investment grade after the two roll periods in the graph. We attribute this rally to the value of time decay (or the roll-down) in high yield, given that default rates on CDX have been much lower than implied by market spreads. This phenomenon has some similarity to the super senior and equity relationship for investment grade tranches. Consider the following relationship where we imply the spread for the 15-100% portion of the HY CDX index by selling protection on the entire index and buying protection on 0-15% (see Exhibit 4). The implied premium on 15-100% ranges from 63 bp to 135 bp for the three series of the index, which shows the value of the roll-down in maturity between the three indices. In Exhibit 4, we also compare the implied senior position spreads to those actually quoted in the market, which are generally tighter. This should not come as a surprise as the premium cash flows for the synthetic position will be marginally reduced by the first default, while the spreads on the actual tranches are only exposed to losses in excess of their attachment points.
Please see additional important disclosures at the end of this report.
277
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 39
exhibit 4
Maturity Net Upfront Payment (Mid)*
Understanding High Yield Tranche Roll-down HY CDX 3 12/20/2009
HY CDX 4 6/20/2010
HY CDX 5 12/20/2010
11.65%
11.18%
11.74%
Index Strike Spread
375
360
395
Spread Funded by Upfront (at Libor)
312
270
260
Implied 15-100% Spread
63
90
135
Weighted Average of Quoted (Mid)
33
51
82
Spread lost after first default
3.8
3.8
3.8
Note: This analysis shows the implied premium on a 15-100% tranche of HY CDX by selling protection on the index and buying protection on the 0-15% tranche *Net Upfront payment from selling index protection and buying 0-15% protection, nominally weighted. Source: Morgan Stanley
PORTFOLIO COMPOSITION – POSITIONING FOR INVESTMENT GRADE INDEX DIFFERENCES
Within the investment grade index tranche space, we find some simple relationships interesting. Inter-series trades (between CDX series 3, 4 and 5) involve both maturity and portfolio composition mismatches. Given that tranches are exposed only to aggregate defaults in the portfolio over fixed time horizons, these marginal changes in portfolio composition and timing of defaults can be difficult to quantify. In examining these relative opportunities among series we often find ourselves making these kind of trade-offs. Clearly the value of time and portfolio composition can have very different impacts depending on where in the capital structure we are. While quantifying may be difficult there is no arguing that some risks just feel cheap relative to others even if we can’t say anything about absolute pricing. In Exhibit 5, we show the difference in 5-year 3-7% tranche spreads for CDX 5 versus CDX 4 compared to the difference in average spreads for the names that are not common to both portfolios. The idea of the analysis is to capture the impact of the portfolio differences on the 3-7% tranches, generally assuming that the value of the 6 month maturity difference is reasonably stable. At several times, the spread differential has moved significantly without a meaningful change in the spreads of the marginal names, although today the relationship seems more in line. In our view, this is a good relationship to follow, as the market likely does not focus on it so closely. Ironically, the relationship in the 7-year version of this tranche seems more stable recently, which likely demonstrates how the riskier tranche (7-year) is more influenced by the spread moves in the “marginal credits.”
278
exhibit 5
CDX 4 versus 5 – 5 Yr 3-7% Spread Differentials
Non Common Name Diff
Non Common Name Spread Differential 3-7% Spread Diff
190
Buy the basis here?
180
3-7% Spread Diff 25 20
170
15
160 10 150 5
140
0
130 120 20-Sep-05
4-Oct-05
18-Oct-05
1-Nov-05
-5 15-Nov-05
Source: Morgan Stanley
exhibit 6
CDX 4 versus 5 – 7 Yr 3-7% Spread Differentials
Non Common Name Diff 190
Non Common Name Spread Differential 7yr 3-7% Spread Diff
3-7% Spread Diff 100 90
180
80 170
70 60
160
50 150
40 30
140
20 130 120 20-Sep-05
10 04-Oct-05
18-Oct-05
01-Nov-05
0 15-Nov-05
Source: Morgan Stanley
WHAT ARE THE TRADES TO DO?
As we have described in detail in this chapter, our approach is to focus on the big picture behind the relative value. In the term structure space, we continue to favor selling protection in 10-year 3-7% versus 5-year 0-7%, given risk allocation to the last 5 years of the trade (see “Structured Credit Ideas,” October 18, 2005), and we continue to find the index “basis” in 5-year 3-7% interesting, recommending that investors buy protection in series 3 and sell it in series 4 or 5. (see Chapter 53). Particularly in series 4 versus 3, the recent spread of about 40 bp seems excessive even given the increased auto exposure.
Please see additional important disclosures at the end of this report.
279
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 40
Tranche Forwards – The Next Leg
January 6, 2006
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Pinar Onur As the derivatives culture gains momentum within US credit markets, ideas like selling 5-year into 2-year tranche protection – derivatives-speak for forward starting investment strategies – have emerged as interesting ways to both play the credit cycle and identify relative value among a broad base of liquid instruments (see “Structured Credit Ideas,” October 18, 2005, and Chapter 47 for our earlier thoughts and ideas). Derivatives 101 (or maybe it is 301) tells us that if implied forward rates consistently disagree with what is actually realized when time passes, then there is some amount of forward risk premium (negative or positive) embedded in market pricing. In the interest rate world, positioning for this forward risk premium is relatively simple, but in credit markets, any perceived forward premium may actually be fully attributable to shifts in credit risk, as well. In this chapter, we look back on some work done by our interest rate strategy team that demonstrates the amount of positive (although shrinking) forward risk premium in short-term rates and attempt to apply this logic to the credit markets. In particular, we examine forward premium relationships in DJ CDX 4 and 5 tranches and conclude that the 5-year into 2-year 3-7% in CDX 4 seems attractively priced relative to CDX 5, based both on correlation and forward risk premiums. A WORD ON TERMINOLOGY
Two aspects of forward trades make them confusing. First, implementing such trades generally requires multiple legs, even though the terminology used can make it sound like there is only one leg. Second, as with much in derivatives, the terminology itself can also be confusing because different jargon is used to describe the same idea. Selling 5-year into 2-year 3-7% protection means selling 2-year 3-7% protection 5 years from now, which is equivalent to buying 5-year 3-7% protection and selling 7year 3-7% protection on the same index or portfolio.
280
FORWARD PREMIUMS IN INTEREST RATES
Our interest rate strategy team has published interesting analysis comparing 3-month into 3-month forward rates with the actual three-month rates three months hence (see “Interest Rate Strategy: Resolving the Conundrum,” November 14, 2005). In general, they find that there is positive risk premium over time, meaning that investors receive additional yield to lock in a 3-month rate through a 3-month forward agreement, compared to actual rates three months forward. This phenomenon was true through most time periods since 1994 in the US, although both the 1994 time period (where markets underestimated Fed tightening) and 2005 (where economic growth was perhaps stronger than expected) forward risk premiums were negative. In non-US markets, forward risk premiums have also been positive, in general. exhibit 1
Five Years Forward – BBB Default Rates Jump, But Bond Spreads Don’t
1970-1995
Average 5 Year Cumulative Default Rate 1.92%
Average 5 Into 5 Year Cumulative Default Rate 2.94%
1989-1995
0.85%
2.16%
1989-2005
Average 5 Year Spread 158bp
Average Implied 5 into 5 Year Spread 157bp
1989-1995
141bp
132bp
Note: Spreads are over Treasuries. Source: Morgan Stanley, Moody’s
However, 3-month instruments are not great analogs for liquid credit instruments, so we are tempted to extend our interest rate strategy team’s analysis to longer maturities. We examined the difference between the implied 5-year into 5-year forward swap rate and the actual 5-year swap rate realized five years later. On average, the forward has overestimated the realized by about 240 bp. However, our interest rate strategists accurately point out that we are looking at this relationship in the context of a period of generally declining rates. In an attempt to address the issue, we adjusted our forward premium for differences in the real short rate at the time of the forward reading and the actual 5-year spot reading. This very rough adjustment reduces the implied forward premium to roughly 160 bp, but it remains solidly positive. While we recognize that a clever economist could drive an aircraft carrier through the holes in this analysis, it serves to illustrate a point. It seems reasonable to assume that there is some positive premium in the interest rate markets for locking in forward rates.
Please see additional important disclosures at the end of this report.
281
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 40
Is this forward rate risk premium analysis applicable to credit? A few points are worth making, in addition to our previous assumption that, all things being equal, there should be some risk premium for locking in forward prices. First, given the youth of the market, it will likely be difficult to derive what a fair premium looks like (this is even hard to do in interest rates). Second, given the risks of default and credit quality migration (mainly downward for investment grade credits, but both up and down for high yield), we would expect more risk premium for credit-risky forward instruments. FORWARD PREMIUMS IN INVESTMENT GRADE CREDIT
When the default swap term structures first started to gain transparency and liquidity a couple of years back, we discussed how default risk can be distributed through time and argued for how CDS curves ought to reflect this phenomenon (see “Getting Short the Long End,” October 10, 2003). In Exhibit 1, we show both the average 5-year default rates and the 5-year default rates 5 years forward for BBB credits for cohorts from 1970 to 1995. Comparing these losses to actual spreads gives us some interesting insights. First, 10year default risk is more concentrated in the last five years than in the first five years, which is a result of the negative ratings transition phenomenon of investment grade credit. Second, cash bond spreads have not generally reflected this trend in the past. We offer the observation that in today’s market, the basis between cash bonds and CDS grows more positive the further we go out in maturity, indicating that derivatives markets may better reflect this difference in risk over time than cash markets do. As with interest rate markets, it is difficult to make any absolute statements, given the lack of robust data. FORWARD PREMIUMS IN CREDIT TRANCHES
Tranche pricing can be affected by both of these of issues in meaningful ways. For instance, the fact that high yield tranches price with a much flatter correlation curve than investment grade tranches can at least partially be explained by the fact that a much larger proportion of high yield spreads represents default risk, as well as the fact that, at times, investors may not be sufficiently compensated for that risk. Part of the correlation skew can also be explained by the difference in the allocation of the default risk and the risk premium in spreads.
282
exhibit 2
Tranche 0-3%
Tranche Implied Forward Spreads, CDX 4 vs. CDX 5 CDX 4 Implied Forward Spreads 5 Year into 2 Year Spread 7 Year into 3 Year Spread 2,495 2,032
3-7%
828
7-10%
105
350
10-15%
68
153
15-30%
16
42
Tranche 0-3%
3,759
CDX 5 Implied Forward Spreads 5 Year into 2 Year Spread 7 Year into 3 Year Spread 2,218 1,643
3-7%
618
7-10%
111
369
10-15%
76
159
15-30%
18
44
Tranche 0-3% 3-7%
3,340
Difference (Series 4 Minus Series 5) 5 Year into 2 Year Spread 7 Year into 3 Year Spread 277 389 210
419
7-10%
(6)
(19)
10-15%
(7)
(6)
15-30%
(2)
(2)
Source: Morgan Stanley
While it is extremely difficult to judge what a “fair” amount of risk premium would be for forward positions in tranches, we still find opportunities in the space. For those who find the above forward risk premium argument appealing, the relative cheapness should be evident using simpler metrics, as well. To illustrate this, we examine the pricing of CDX IG 4 and 5 tranches across the curve. In Exhibit 2, we show the implied forward spreads for tranches across the 5-, 7- and 10-year maturities, based on recent market prices. There is a remarkable consistency between the two series in terms of the forward (and spot) prices charged for various senior tranches across the two series, while the pricing on junior tranches have some interesting implications. Portfolio composition (auto risk versus LBO risk), inherent differences in spread dispersion and maturity differences can all contribute to the price differences between the indices, although the impact is likely not consistent through time and across the capital structure.
Please see additional important disclosures at the end of this report.
283
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 40
The tranche that stands out to us is the 3-7% in CDX 4, 5-year into 2-year. Relative to CDX 5, it is 210 bp wider. Additionally, we find that the pick-up from CDX 5 to CDX 4 on 3-7% 7 years into 3 years provides more spread than pick-up on the 0-3% tranche for 5 years into 2 years. Both of these seem to offer opportunity to capture value. However, isolating these risks requires four legs, so investors may consider using these as indicators of relative value for the spot tranches, as well. This argues for selling 37% 7- and 10-year protection in CDX 4 versus CDX 5, as well as selling protection in 7-year 0-3% in CDX 5 versus CDX 4. However, the latter requires an upfront payment, which amounts to about one-third of the potential positive cashflow in the case of a single default in the marginal CDX 4 names (see Exhibit 4). ARE THERE OTHER WAYS TO IDENTIFY THE CHEAPNESS?
While we would argue that the above forward spread analysis is complicated by almost any measure, thankfully we find consistent relative value themes when we look at implied correlation (see Exhibit 3). The base correlation skew for 7 year 3-7% in CDX 4 is lower (cheaper) than for CDX 5 (despite the 6-month-shorter maturity). In 5-years, it is the other way around, with CDX 4 base correlation skew higher in CDX 4 (richer) than in CDX 5. The difference is 3.7 correlation skew points, which we find appealing, but it requires the box trade to implement. exhibit 3 Tranche 7 Year 3-7%
CDX 4 13.9%
CDX 5 15.9%
5 Year 3-7%
16.8%
15.1%
2.9%
-0.8%
Difference Source: Morgan Stanley
284
Correlation Curves Are Inverted
WHY THE CHEAPNESS? THE PRICE OF AUTO RISK
Thankfully, there is a very fundamental reason why the relative value described by the above derivatives-speak actually exists. The answer, in a nutshell, is the higher auto risk in CDX 4 relative to CDX 5 (see Exhibit 4). While correlation models will determine auto risk at face value (i.e., auto spreads), the tranche market is requiring even more compensation for 3-7% 7-year type investments, because the ultimate direction of the three names in question is the difference between this tranche being more mezzanine-like than equity-like. Yet, we feel that at today’s spreads, the additional compensation is more than adequate. exhibit 4
Portfolio Composition IG CDX 4 VS 5
In CDX 4 But Not CDX 5 Eastman Kodak Company
In CDX 5 But Not CDX 4 The Gap Inc.
Ford Motor Credit Company
IAC InterActive Corp.
General Motors Acceptance Corporation
Knight Ridder Inc.
Kerr-McGee Corporation
Limited Brands Inc.
Lear Corporation
Marsh & McLennan Cos. Inc.
Liberty Media Corporation
Radio Shack Corporation
Maytag Corporation
Sabre Holdings Corp.
MBNA Corporation
Sara Lee Corp
Sears Roebuck Acceptance Corp.
Toll Bros. Inc.
Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
285
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 41
Self-Managed Synthetics – Seeking Stability
February 24, 2006
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Pinar Onur Adjustable subordination technology, not a term that turns many heads, has played an important role in the near renaissance of managed synthetic CDOs over the past year. While managed synthetics are nothing new, they lost their appeal as single-tranche transactions (including index tranches) gained considerable popularity over the past few years. One offshoot of these two trends is growth in popularity of self-managed single tranche synthetics, which as the name suggests, allow investors to take exposure in tranche form to a portfolio that he or she actually manages. Typically, no other exposure in this portfolio would be distributed by a dealer. The technology, which effectively serves as a protocol between a dealer and a manager, allows managers to swap credits in an underlying portfolio while dealers adjust their hedges simultaneously. The resulting economics are translated into increased or reduced subordination for the tranche in question. While one could argue that the process is a bit more cumbersome than simply allowing a manager to trade credits without any exchange of hedges, the latter requires a fully distributed synthetic CDO, which today has become far less common. INVESTOR MOTIVATIONS
Why would one choose a self-managed structure to gain exposure to credits, versus the alternatives such as a third party managed transaction, a static transaction, or simply a straight single-name portfolio? As we see it, the motivation for a self-managed transaction is leveraging one’s own credit-picking skills (this is obvious), but not necessarily for immediate economic reasons (this is not so obvious). Both ratings and income stability are key motivating factors, for which an investor is giving up realization of P&L. While one could lever management skills by buying a tranche of a public CDO they manage, there is clearly a lot of overhead in this process, making self-managed much simpler. We have found that banks and insurers have been the most likely to embrace selfmanaged structures, given the importance of income and ratings stability and the ability to leverage internal credit-picking skills. And somewhat ironically, self-managed structures have also been popular in synthetic buckets of managed cash CDOs. SELF-MANAGED TECHNOLOGY
While potentially a bit cumbersome, market practices for self-managed structures are fairly simple. From the perspective of the investor, implementing a change in the portfolio is equivalent to unwinding an old single-tranche trade and putting on a new one, but without any payments or realization of gains and losses. The premium and width of the tranche will remain the same, but the attachment point will move up or down to reflect the mark-to-market impact of swapping a credit in the portfolio. Replacing a wider trading name with a tighter one would result in a loss of
286
subordination; the opposite would result in a gain in subordination. In addition to the tranche mark to market, the loss in subordination would also reflect any cost associated with hedging required by the dealer as a result of the new portfolio composition. exhibit 1
Attach Detach Tranche Premium Market Value Attach Detach Spread Market Value
Initial 3.00% 7.00% 99 bp 0.00% 7.00% 10.00% 22 bp 0.00%
Translating a Trading Loss to a Subordination Shift Single Credit Spread Widens to 300 3.00% 7.00% 99 bp -0.54% 7.00% 10.00% 22 bp -0.13%
Replace Credit 2.865% 6.865% 99 bp -0.54% 6.810% 9.810% 22 bp -0.13%
Change in Subordination -0.135%
-0.190%
Source: Morgan Stanley
In Exhibit 1, we show an example where the manager/investor sells a credit that widened 300 bp (from its initial premium of 44 bp) and buys a new credit at 44 bp. When the credit widens by 300 bp the price impact on the 3-7% tranche (assuming CDX 5) amounts to a -0.54% mark-to-market loss (for 7-10% the loss is 0.13%). The investor essentially swaps this position for a tranche on the new portfolio with the same market value (we ignore any hedging costs in this illustration). Because the new portfolio is marginally higher quality (one credit trades 300 bp tighter), this new tranche would have to attach lower given it has the same spread and thickness as the old one. In this case, that loss in subordination amounts to 0.135% for 3-7% and 0.19% for 7-10%. One key point is that the 300 bp widening amounts to about a 12% price impact, which is roughly 20% of the loss on a 40% recovery default. SELF-MANAGED VS. STATIC TRANCHES
One of the most obvious advantages of self-managed structures is that once a credit begins a downward spiral to default, holders of self-managed CDOs would not have to wait until the actual default to get out of the risk. Investors essentially have the opportunity to identify a deteriorating credit, remove the risk from the portfolio and recognize a smaller loss than they could otherwise. Similarly, investors could build subordination by trading out of names that have tightened dramatically. Trading a widened name out of the underlying portfolio would likely cause a loss in subordination in the tranche (as illustrated above). However, the size of this loss in subordination can be much smaller than the impact of an actual default, depending on how quickly a credit is recognized by the investor. As Exhibit 1 highlights, the impact of removing the name on the 3-7% tranche subordination level amounts to 30% of the potential 0.48% subordination loss in the case of a default. Obviously, the sooner the credit is identified the greater the subordination savings versus a default, so credit skills provide leverage in this context. PLAYING DEFENSE
The natural extension of this notion is the idea that a self-managed transaction can withstand a larger number of “mistakes” in the portfolio without affecting ratings compared
Please see additional important disclosures at the end of this report.
287
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 41
to a static transaction. In Exhibit 2, we show the number of defaults that can be withstood by both a managed and static transaction, assuming the manager/investor is able to remove a credit and suffer only one-third of the loss in subordination from an actual default for an investment grade portfolio. We point out that there are a significant number of opportunities to correct portfolio “mistakes” in a managed structure. This is particularly true when we consider this relative to reasonable default and transition rates for investment grade credits. exhibit 2
Attachment Point 7.00% 5.00% 4.50% 3.50%
Number of Immediate Events before Whole Letter Downgrade (CDX Series 5) Estimated Initial S&P Rating AAA AA A BBB
Static Transaction 4 1 2 3
Self-managed Transaction 12 3 6 9
Note: Assumes managers remove default candidates with 33% of full default subordination loss. Source: Morgan Stanley
The risk in these strategies is that the investor/manager chooses to remove names from the portfolio at a subordination loss and that the name then recovers to trade at near original spread levels. This risk is directly related to the credit-picking skills of the investor and presumably, investors who are the least likely to make these erroneous portfolio changes are those that will be most attracted to self-managed structures. One can play offense as well using self-managed strategies, building subordination as wide trading names tighten, but we still see defense as the most common tactic. THE ECONOMICS – STATIC STRUCTURES WITH DELTA HEDGES
Self-managed structures can be thought of as being similar to trading a static tranche transaction while dynamically managing single name deltas, but there are subtle differences in the economics of the two. First, in self-managed transactions, when a credit is added or removed from the portfolio, the delta is selected at a specific point in time and the tranche is essentially marked-tomarket. Any further impact on the tranche for the name in question is effectively eliminated from the transaction. The same cannot be said for static tranches and dynamic delta hedging. First, deltas for a given credit can change based on the spread of the specific credit, the other spreads in the portfolio, the current pricing regime for the particular part of the capital structure (i.e., implied correlation), as well as the passage of time. Any dynamic strategy still has the risk that as the delta changes the hedges do not match reality. Second, even if delta hedges are in place, the rating for a given tranche transaction will be affected by the original underlying portfolio results regardless of any economic offsets achieved through hedging strategies. Even if investors have perfectly hedged their economic risk, they face the risk that the original transaction is downgraded. For investors who base their decisions on cost of capital (or regulatory capital) or who have specific investment guidelines, this ratings migration risk is significant and can be the difference between profitability and loss. In fact, it can be more important than the economic risk hedged via delta hedges.
288
SELF-MANAGED VS. STRAIGHT SINGLE-NAMES
While comparing a self-managed tranche to a static one is natural, another worthwhile effort is to think of such tranches relative to a straight single-name portfolio. As a motivating anecdote, a key performance theme last year was that auto-induced idiosyncratic risk was largely responsible for an underperforming single-name market. In the tranche world, equity tranches were the clear loser, while the junior-most mezzanine tranches were the winners (see Exhibit 3). exhibit 3
CDX vs. Tranches Performance
2005 Returns Investment Grade Excess Return GM + Ford Contribution
-0.60% -0.46%
IG 5 Yr Index & Tranches CDX Index
0.30%
3-7%
3.76%
7-10%
1.57%
10-15%
0.43%
15-30%
0.12%
Source: Morgan Stanley
As it turns out, a static junior mezzanine tranche (3% subordination) was a good enough way to escape the auto-pain last year, given its relative performance to everything else. But in a riskier credit environment, a junior mezzanine tranche could be the weak performer, depending on how much idiosyncratic risk there is. This is the motivation for a credit picker to choose a self-managed structure as a leveraged bet on one’s own credit skills. There can be huge performance differences in junior mezzanine tranches with the right credits vs. the wrong ones. Compared to single-name portfolio investors, self-managed tranche investors are looking for opportunities to “spend” or “build” subordination so that idiosyncratic risk does not affect them. While ratings may be less important in this context, successfully building subordination may result in more stable ratings relative to the average of the underlying portfolio, which can be important to many (see above). Furthermore, since the coupon from such a structure is fixed, there can be some interesting asset liability impacts versus using a portfolio where coupons change as the portfolio is traded. A TOOL IN THE TOOLKIT
While we find self-managed technology interesting for those who want to leverage credit-picking skills and seek ratings and income stability, there are some issues. Trading is clearly more complicated, and we do not see managers/investors being overly active in this context. Liquidity could also take a back seat in a negative credit environment, but it is important to realize that managers/investors are long the option to make substitutions, although they are not required to do so. What we find most appealing is the ability to translate the economics of trading single-names into a form (subordination shifts) that can potentially stabilize ratings and income.
Please see additional important disclosures at the end of this report.
289
Section D
Market Themes
Please see additional important disclosures at the end of this report.
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 42
Swinging a Double-Hedged Sword
June 20, 2003
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Anisha Ambardar In Chapter 9 we wrote about the important correlation and spread relationship in synthetic tranches that we can observe in the market. We received quite a bit of feedback on this topic, generally tilted toward questions about the relationship and where the market is headed. In this chapter, we are focusing on the price sensitivity of synthetic tranches, which we believe is critical to understand before market participants jump into this market sector or otherwise “trade correlation.” In particular, we illustrate the inherent leverage of subordinate tranches, and focus on one important application, namely, the use of tranches to offset the volatility of single-name credit default swaps. This hedging of volatility is an issue for any institution required to mark-to-market their credit derivatives but not their cash exposures (e.g., banks and insurance companies). OBSERVING LEVERAGE
In Exhibit 1 we compare the implied spread sensitivity of two TRACERSSM tranches to the (untranched) TRACERSSM basket over an eight-month time period. The spread values of the tranches are derived from our expected loss model (using a fixed correlation assumption), as we only have recent market prices for TRACERSSM tranches. exhibit 1
The Inherent Leverage in Subordinate Tranches
bp
1,100 Tracers 6 - 10% 10 - 15%
1,000 900 800 700 600 500 400 300 200 100 Oct-02
Dec-02
Jan-03
Mar-03
Apr-03
Jun-03
Source: Morgan Stanley
The key observation in this analysis is the leverage. The TRACERSSM basket rallied approximately 108 bp over this period, while the 6-10% and 10-15% tranches rallied 865 bp and 404 bp, respectively. Based on these relative moves, the inherent price leverage in the two tranches is 7x and 5x, respectively.
292
DON’T GET FOOLED BY THE RATING
There is one more point that we believe is critical to understand. The ultimate default risk of a tranche (when issued as a funded note) may be low enough to get a high rating, but one should realize that agencies are rating the risk of default and the severity afterwards. The price volatility of an instrument may not be a concern in assigning or maintaining a rating. While one could argue that traditional credit investors should be used to this phenomenon in the single-name credit markets, given the cyclical nature of many credits, the leverage in tranches may introduce a higher level of price volatility compared to economic cycles. In short, investors in synthetic tranches must be aware of these differences in price sensitivity and of the fact that credit rating alone can be a misleading measure of price sensitivity. VOLATILITY OF A HEDGED POSITION
Banks and insurance companies who are credit derivatives users suffer from income statement/earnings volatility because of FAS 133 issues. In particular, many of their cash credit investments are not marked-to-market, while their derivative investments or hedges require mark-to-market treatment. Ironically, this situation creates a disincentive for those seeking to hedge credit exposures, based purely on accounting rules, assuming they are unable to meet the requirements for hedge accounting. We consider banks as a specific example. Many banks have been buyers of protection to partially hedge their corporate credit exposure in loans. As their loans are not marked to market, the credit default swaps positions can add earnings volatility although, in fact, they are simply hedges for concentrated credit exposures. Additionally, banks must contend with the fact that credit default swaps are higher beta instruments (because of higher volatility and a propensity to reinforce trends) and may indeed move more dramatically than cash instruments in a credit improving environment. This implies that there would still be a mark-to-market impact, even if the cash instruments were marked. Banks, for the most part, have to live with these problems unless they are willing to offset their long protection positions with long credit positions (via credit derivatives to get the same accounting treatment). One simple solution is to offset the purchase of protection on some names with the sale of protection on others, but this process adds credit risk to portfolios, and requires banks to take name-specific (or idiosyncratic) risk. This risk is precisely what banks are trying to shed in the first place, so this long protection/short protection strategy on single names does not necessarily help. GETTING LONG DIVERSIFIED CREDIT RISK
Another method banks can use to get long credit risk through derivatives is to consider diversified credit portfolios. Clearly, getting long credit through a liquid basket (such as synthetic TRACERSSM) is one way to gain diversified long credit exposure, but there may not be enough spread in the basket to offset the premium paid out on the long protection positions (although full premium offset may not be a goal). Further, there is no first-loss protection in such a diversified basket, which can result in additional (or unexpected) idiosyncratic risk in the bank credit portfolio.
Please see additional important disclosures at the end of this report.
293
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 42
A second approach is to consider selling protection on synthetic tranches, which can have a similar impact from a carry perspective, but with differing sensitivity to moves in credit spreads depending on the exact tranche. Yet, the inherent leverage in these tranches could be the second blade of a double-“hedged” sword used to hedge both idiosyncratic credit exposure and financial statement volatility. WHAT IS A GOOD HEDGE?
We construct a hypothetical example. Using the benefit of hindsight, suppose a bank (in April 2002) wanted to hedge what turned out to be the 10 largest tightening and 10 largest widening names in the 100-name TRACERSSM for a volatile one-year period (until April 2003). If we assume the purchase of $10 million of protection on each name, the first data column of Exhibit 2 describes this portfolio ($200 million notional with an average premium outlay of 144 bp). exhibit 2
Hypothetical Portfolio and Hedges (April 2002)
Original Portfolio Buy
Untranched Sell
7-10% Sell
10-15% Sell
Notional
200,000,000
194,000,000
38,000,000
50,000,000
Initial Premium
(144)
91
339
Net Premium
(144)
(56)
(80)
Protection
174 (101)
Source: Morgan Stanley
We then construct three packages, meant to serve as hedges for this portfolio of protection. The first hedge is simply selling protection on the full TRACERSSM basket, using a risk (or spread PV01) neutral notional amount. The second package is to sell protection on the 7-10% TRACERSSM tranche, and the third package includes the sale of protection on the 10-15% tranche (in both cases with a PV01-neutral notional amount). The net premium on all of these packages is negative (but still better than just buying protection on the 20 names). exhibit 3
Original Portfolio with 7-10% Tranche Hedge -1.73%
Original Portfolio with 10-15% Tranche Hedge -0.83%
-0.22%
-1.82%
0.06%
0.26%
-0.06%
0.63%
1.19%
0.83%
1.33%
0.52%
1.34%
0.74%
1.50%
0.66%
Original Portfolio Alone 0.49%
Original Portfolio with SM TRACERS Hedge -0.91%
10/01/02
3.60%
01/02/03
1.45%
04/01/03 Std Dev
07/01/02
Source: Morgan Stanley
294
Performance and Volatility Comparison – Which Hedge Is Best?
As the bank’s goal is to reduce mark-to-market volatility, we analyze the performance of the protection purchase with the three packages we described above (see Exhibit 3). Over the subsequent four quarters, the estimated P/L (excluding the impact of carry) on the 20-name portfolio of protection is shown with values ranging from 0.49% to 3.60%. The standard deviation of these four quarterly values is 1.34%. The first package reduces the volatility of the P/L to 0.74% and moves the quarterly P/L percentage closer to zero in three of the four quarters. The second package (with 7-10% tranches) does a worse job by increasing the overall volatility and moving the P/L percentage further away from zero in two of the four quarters. The last hedge performs the best, with the overall volatility at 0.66% and P/L percentage changes closer to zero in three of the four quarters. WHAT IS THE RIGHT ANSWER?
The hedge that will perform the best (i.e., near zero P/L volatility) is well correlated to the original portfolio of protection, but does not introduce idiosyncratic risk into the whole process. An untranched basket has the advantage of simplicity and liquidity, but can introduce idiosyncratic risk because of the lack of first-loss protection. Tranches can be used to solve this problem (and can be higher yielding), but the inherent leverage and carry impact must be managed carefully. We point out that if we examine the results in Exhibit 3 for hedges weighted to be carry neutral rather than PV01 neutral, the volatility of the P/L is markedly increased for both the tranched and untranched hedges. Both approaches can reduce overall P/L volatility and premium outlay, so we encourage institutions with mark-to-market accounting issues on hedges to think about the idea of “hedging” their mark-to-market volatility. Tranches, more specifically, can be structured to provide exposure to aggregate credit losses in a portfolio with subordination providing a buffer against idiosyncratic default risk.
Please see additional important disclosures at the end of this report.
295
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 43
Mezzing with the Markets
September 17, 2004
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Viktor Hjort An age-old battle in financial markets involves the seesaw between technicals and fundamentals, which is very much front and center in today’s credit markets, on both sides of the Atlantic. Fundamentals appear strong currently, given the nature of most corporate balance sheets, but the warning flags have been raised as many companies look to execute shareholder-friendly activities rather than keep leverage at suboptimal levels. Spreads seemed like they were priced to perfection just a few weeks ago, but the ominous structured credit bid has almost driven us into a new regime, making the battle even more acute. As we see it, the structured credit bid seems to have some strong legs in the current environment. With rates remaining low and equity markets showing uninteresting returns, the reach-for-yield phenomenon is naturally pushing investors toward products that offer higher yields. The current fundamentals in the credit markets make levered credit a natural choice because default risk seems to be low, independent of the current level of spreads.
© 2004 Morgan Stanley
While investors do not need to look beyond the single name markets to get a sense of how much impact the structured credit bid is having on valuations, the relative pricing of benchmark tranches is in fact much more enlightening. At current spread levels, junior mezzanine tranches feel like out-of-the-money options on default, and that phenomenon, combined with the demand for yield, has pushed them significantly tighter. Implied correlation values reveal some of this recent richening as well, but there are other, more subtle factors helping to drive spreads tighter.
296
Has the market gone too far? From a default risk perspective, taking credit risk continues to feel fairly attractive, given the healthy state of corporate balance sheets today. Those who go long equity and mezzanine tranches are, in effect, expressing this view, so it may be difficult to argue with that logic. However, there is a price for everything, and we explore the potential sensitivity of some tranche positions to events that are not priced into the market, such as a spread-widening environment or a random credit event. THE STRUCTURED CREDIT BID
The impact of the structured credit bid has been felt by everyone in the investment grade markets over the past few weeks, with already tight default swap premiums tightening 12 bp in the US and 11 bp in Europe. Benchmark mezzanine tranches are an even better indicator, with rallies of 100 bp and 75 bp for the US and Europe, respectively, using 3% attachment points since the middle of August. Models aside for the moment, from an optical perspective, junior mezzanine notes feel quite rich, particularly an 85 bp spread on 3-6% in Europe. exhibit 1
Mezzanine Tranches vs. the Market – Europe
iTraxx Europe Spread
3-6% Tranche Spread 450
65
400
60
350
55
300
50
250
45
200
40
150 100 50 0 Jul-03
35
Roll from Trac-x to ITraxx
30
3-6% ITRAXX/TRAC-X 5yr mid
Oct-03
Dec-03
25
Feb-04
Apr-04
20 Sep-04
Jul-04
Source: Morgan Stanley
exhibit 2
Mezzanine Tranches vs. the Market – US
CDX NA Spread 75
3-7% Tranche Spread 500 3-7% Tranche (Left Axis) CDX NA (Right Axis) 450
Roll from Trac-x Series 2 to CDX 70
400 65 350 60 300 55
250
200 Oct-03
50 Dec-03
Feb-04
Apr-04
Jun-04
Sep-04
Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
297
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 43
How deep is this bid? In the old days, the CDO bid had reasonable impacts on investment grade and high yield markets, but the size and scope of the current bid is quite a bit more substantial, in our view. It certainly involves traditional buyers of structured credit products, including European and US institutions, but what is new this time around are both retail investors in Europe and institutional investors in Asia (outside of Japan). Perhaps one reason why the flows are occurring when they are is the general frustration with the alternatives, including the equity markets and interest rates that are not rising as much as many had hoped. Effectively, investors are expressing the view that economic growth will be unexciting but still positive, while defaults remain low. Is anyone leaning against the wind? We have seen some minimal amount of structured credit unwinding in the face of very tight spreads, but that flow is much smaller. Most who locked in the bigger coupons from years past appear to be quite happy continuing to clip the coupons, even if there is some mark-to-market risk going forward. However, we do note that there has been structured credit-related net selling of credit risk in the past that has impacted the markets (the 3-100% tranche trades, for example; see “Equity Indicators – Is the Tail Wagging the Dog?” February 27, 2004). YOU CAN SEE IT IN MEZZ PRICES
If we dig deeper into the junior mezzanine activity, we can attribute the tightening to three factors: the move in the underlying market, the technical flow (i.e., correlation), and a change in the shape of the distribution of the index. The first two factors are both substantial (as shown in Exhibits 1 and 2) and obviously somewhat interrelated. The base correlation move was approximately 3 points (over the past five weeks), making the base correlation curve now worth nearly 11 points between the junior mezzanine and equity tranches in the US. exhibit 3
Redistribution of Credit Spreads Makes Junior Mezzanine Less Risky
35 30
2/19/2004 9/15/2004
25 20 15 10 5
200 -
190 - 200
180 - 190
170 - 180
160 - 170
150 - 160
140 - 150
130 - 140
120 - 130
110 - 120
100 - 110
80 - 90
90 - 100
70 - 80
60 - 70
50 - 60
40 - 50
30 - 40
20 - 30
0 - 10
10 - 20
0
Source: Morgan Stanley
The market and correlation moves clearly overshadow more subtle impacts, but investors can gain some insight if they focus on these effects as well. Model-based valuation of mezzanine and equity tranches can be very sensitive to the size and shape of the tails of the credit distribution because this tail is, in effect, what increases the probability of an out-of-the-money option becoming in-the-money. Six months ago, we
298
took the somewhat wider tails that we saw in equity indicators (Moody’s KMV) and considered the impact they would have on the benchmark tranches. At that point, we found that the longer tails of the equity market were beneficial to attachment points of 3% and above, if in fact the equity indicators eventually wound up influencing credit spread distributions (see Chapter 27). The distribution of credit spreads within CDX has indeed reshaped since then, in a way that is beneficial to 3-7% type tranches (see Exhibit 3). We note, however, that the European indices and tranches did not experience a similar change. 3-6% EUROPE VS. 7-10% US
A trade idea we find interesting is one that plays against the absolute level of spreads that influence tranche pricing so much. Clearly, a random default can come from anywhere, but the cost of buying this random default option in Europe is much less expensive than in the US. As such, a long protection position in the 3-6% European tranche versus a short protection position in the US 7-10% tranche would benefit from the random default in Europe while not being penalized as much by a similar event in the US. Additionally, this is a flat carry trade. We approximate the spread move caused by a loss in subordination resulting from a default of the widest credit in the 3-6% iTraxx tranche to be roughly 2.5 times that of the 7-10 CDX tranche (55 bp versus 22 bp). Additionally, the trade has some interesting spread convexity and roll-down or time decay behavior. In Exhibit 4, we have shown the performance of both tranches for a variety of spread scenarios. To normalize the spread movement in Europe versus the US, we assume the iTraxx/CDX basis will widen with spreads subject to a maximum of 25 bp. Under these assumptions, the package benefits from generic spread widening, with potential further upside for scenarios where the basis tightens or simply widens less. exhibit 4
Buy 3-6% Protection in Europe, Sell 7-10% in the US
Tranche Price Impact 20% 3-6% iTraxx (Long Protection) 15%
7-10% CDX (Short Protection) Flat Notional Package
10% 5% 0% -5% -10% -15% 30
50
70 CDX Spread
90
110
Source: Morgan Stanley
WHICH WAY DO WE GO?
For all of the reasons cited, we feel that the structured credit bid is strong and not necessarily going away any time soon. Furthermore, near-term default risk continues to seem attractive to us, even if spread levels do not. For those investors who are inclined to position against the technicals, the 3-6% vs. 7-10% idea takes advantage of the hurricane force winds blowing across the Atlantic today.
Please see additional important disclosures at the end of this report.
299
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 44
Insurance Reshapes the Risk Profile
October 22, 2004
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Ajit Kumar, CFA In a market dominated by technicals of late, the fundamental repricing of credit risk in the insurance sector serves as an important wake-up call for credit investors. It is not common to see a significant increase in credit volatility overnight among tight trading names in any credit environment, and the resulting spread move in the insurance sector has effectively reshaped the investment grade risk profile. This action, in turn, has some interesting implications for the structured credit market, as tranches with varying seniority have exposure to different parts of this reshaped risk distribution.
© 2004 Morgan Stanley
Tight trading names moving somewhat wider generally impact senior tranches, while tight or even average credits moving significantly wider impact junior mezzanine and first-loss tranches, depending on the size of the move. The net effect of the insurance repricing in the structured credit market to date has been a bit of both, although for portfolios including the most stressed insurance credits, the thickening of the righthand tail (risky credits) has had the biggest pricing impact on subordinate tranches. The tranched CDX market avoided the extreme scenario, as the two most stressed names (Marsh & McLennan and Aon) were not part of the index’s 10-name insurance sector.
300
Interestingly, the blossoming of the CDO-squared market this year reflects the desire of investors to hedge exactly this type of risk, even though it was an “unexpected” event by most measures. RESHAPING THE DISTRIBUTION: IMPACT ON TRANCHES
The point of most structured credit investment strategies is to redistribute credit losses. The process of doing this results in the tranches behaving like options on a portfolio experiencing a specific amount of losses due to default. As such, the shape of the risk distribution within the portfolio influences the pricing of tranches (see Exhibit 1). It should be fairly intuitive to recognize that the length and thickness of the right tail influences the pricing of subordinate tranches, meaning that the bigger the tail, the riskier the equity and mezzanine tranches. Similarly, the shape of the middle to left side of the distribution should influence the more senior notes. As credit risk increases (shifts from left to right), senior notes should become riskier. exhibit 1
Insurance Moves to the Right, Impacting Both Senior and Subordinate Tranches
Number of Credits 30 25 20 15 15-30% 7-15% 3-7% 0-3%
10 5
190 - 200
180 - 190
170 - 180
160 - 170
150 - 160
140 - 150
130 - 140
120 - 130
110 - 120
100 - 110
80 - 90
90 - 100
70 - 80
60 - 70
50 - 60
40 - 50
30 - 40
20 - 30
0 - 10
10 - 20
0
Spread
Source: Morgan Stanley
Two credits in the insurance sector serve as useful examples. The move in AIG’s default swap premium (from 18 bp to 34 bp) increases risk in 15-30% type tranches, while reducing risk in super-seniors (30-100%). On the other hand, Marsh & McLennan (which widened from 30 bp to north of 250 bp) is clearly a right tail event, shifting risk from 15-30% type tranches to 0-3%. exhibit 2
Tranche 0-3%
Change in Single-Name Deltas Shows Risk Redistribution AIG+16 bp 12.4%
MMC+240 bp 61.9%
3-7%
8.0%
14.7%
7-10%
0.6%
-27.7%
10-15%
3.1%
-36.3%
15-30%
43.2%
-45.4%
Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
301
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 44
For investors who like to think of single-name risk in tranches on a delta basis, the resulting changes in these deltas tell the same story, but more numerically (see Exhibit 2). The AIG delta to the 15-30% tranche rose 43.2% (from a very small number), due solely to AIG’s widening. Hypothetically, if Marsh & McLennan (MMC) were part of CDX, its delta in the 15-30% tranche would have dropped 45% and would have risen 61.9% in the 0-3% tranche. If we take this analysis one step forward, we can see the actual pricing impact on tranches (see Exhibit 3). The move in AIG (alone) had a marginally positive impact on the 0-3% tranche, approximately 0.28% in price terms, or about 0.02% on a deltaneutral (to the index) basis. AIG’s impact on the other tranches is even more negligible. The Marsh & McLennan move (alone, if it were part of the index) would have caused a 2.24% drop in the 0-3% tranche (0.16% on a delta-neutral basis). The impact on the 37% tranche would have been negative as well, but for the remaining tranches, it would have been a positive event. exhibit 3
0.05%
Tranche Pricing Impact: Two Opposite Examples
0.00%
-0.05%
AIG+16bps -0.10%
MMC+240bps
-0.15%
-0.20% 0-3%
3-7%
7-10%
10-15%
15-30%
Source: Morgan Stanley
THE IMPACT ON THE LIQUID TRANCHES
While it is likely that almost every insurance name with single-name liquidity would appear in at least a handful of bespoke tranches throughout the market, it is also important to focus on the names that appear in CDX, to get a sense of the impact on benchmark liquidity. There are 10 insurance names in the 125-company index, which experienced an average widening of 29 bp since October 14. Two names (Ace Limited and Aetna) moved from average levels to the right tail, with spread levels above 100 bp. Three others moved from the 40-50 bp zip code to the 70-80 bp neighborhood (XL, Cigna, Hartford), while others moved from fairly tight spreads (30 bp and below) to the 40-50 bp range. As CDX does not include the two most stressed insurance credits (currently), the net impact of the insurance move on the index tranches is somewhat different than it would have been otherwise (see Exhibit 4). Relative to a delta-neutral amount of index protection, the 0-3% tranche outperformed, as the moves in Ace and Aetna into the right tail were more than offset by the risk distribution at lower spread levels. The remaining tranches all underperformed the index, as the net effect was to move risk from the far left tail of the distribution into the middle. If Marsh & McLennan and Aon
302
were part of the CDX index, the net performance would have been flipped, with subordinate tranches underperforming the seniors. exhibit 4
Tranche Index Effect Was More Muted – Given Index Composition
0.10% 0.08% 0.06% 0.04% 0.02% 0.00% -0.02% -0.04% -0.06% 0-3%
3-7%
7-10%
10-15%
15-30%
Source: Morgan Stanley
HOW DID THE MARKET REACT? WATCH THE TECHNICALS
Our analyses above show the impact on the various tranches given the insurance sector spread move, alone. Yet there were some technically driven correlation moves, as well, which provide a sense of market direction. The increase in idiosyncratic risk is not supportive of selling protection on very subordinate tranches. Consequently, early this week, we did see correlation levels fall even further. But we should emphasize that falling correlation levels have been a trend even prior to the insurance news and may be related to hedging in the marketplace ahead of the expected structured credit pipeline of issuance. There was a modest rise in 0-3% correlation over the past two days, reflecting the fact that sellers of protection re-emerged, but, again, we emphasize that this could be related to hedge unwinding. It is also important to note that structured credit exposure to the insurance industry is largely a synthetic phenomenon; the cash CDO market has little insurance exposure, given the focus on high yield debt. THE ARGUMENT FOR SUBORDINATION GAINS SOME TRACTION
There are clearly many ways to lever credit today, but when tranches are involved, there are essentially two types of resulting risk: idiosyncratic and systemic. Selling protection on subordinate tranches (junior mezzanine and equity) is a form of leverage that contains both types of risk, while the risk in senior mezzanine (and more-senior tranches) is less idiosyncratic in nature, albeit systemic. Long/short combinations of the two can isolate idiosyncratic risks and have been a popular trade in the market place. The move in the insurance sector highlights the risk in taking a lot of idiosyncratic exposure, though the tranched index market was spared from the most volatile credits. Subordinate tranche positions in portfolios containing the big movers in insurance would not have performed as well as senior tranches. Interestingly, the CDO-squared market exists exactly for this reason, and is effectively a way of taking levered exposure to tranches that have less idiosyncratic risk themselves. An insurance sector type move is one that would not negatively impact the typical senior mezzanine tranche, provided it is contained to a reasonably small number of credits.
Please see additional important disclosures at the end of this report.
303
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 45
Challenging Correlation Cynicism
December 3, 2004
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Ajit Kumar, CFA Much of the derivatives skepticism across the credit markets from a couple of years ago has subsided, given the tests, cycles and general growth that default swaps have experienced. However, a healthy bout of cynicism has developed in the correlation space today, especially given the enormous impact structured credit is having on credit spreads. This sentiment is especially acute in the high yield community for a variety of reasons including the relative novelty of derivatives in this space and the pain (however distant it may now seem) associated with the high yield CBO experience. We continue to view structured credit instruments as being important credit portfolio management tools that allow investors to isolate, hedge, or even amplify certain characteristics of credit risk. The vast improvements in liquidity, transparency, and standardization have been a key driver of our views, as we have watched these markets evolve since the late 1990s. Correlation cynicism, nevertheless, needs to be addressed.
© 2004 Morgan Stanley
304
Among investors who are less involved in the structured credit markets, the most common question we are asked center on what concerns us most in this space. When risks, flows, and technicals are well balanced, we are generally comforted; when they are not, we begin to get worried. The long/short nature of the correlation market today results in a fair amount of balance and is one of our biggest reasons for comfort. We are also reassured by the transparency in the market, and the alternatives that pricing models provide to a pure ratings-based approach. We are most concerned with the impact an “unexpected” default or stressed sector can have on the levered money in this space. We also worry about the imbalance caused by differences in mark-tomarket treatment among investors, and the resulting flows that could occur. A TWO-SIDED MARKET
The original CDO market was a long-only business, where risk was effectively taken out of the market and then repackaged and redistributed into the hands of different investors. The disparate and fragmented nature of the market made it relatively illiquid as well. When the credit cycle turned, most investors in these products were effectively positioned the same way, and the bid for paper was thin, to say the least. Ratings downgrades brought forced sellers to the market, and mark-to-market valuations began to suffer in a truly negatively convex manner. We draw quite a bit of comfort in realizing that today’s structured credit market is a two-way market, which, simply put, means that there is a winner and a loser for almost any event that impacts these structures. For example, as we have written about repeatedly, investors have used mezzanine and senior tranches as ways of implementing long and short views. However, a derivatives market does not have to be well balanced, and we do worry a bit about crowded trades and accounting differences among investors, which can tilt the balance. THE UNEXPECTED DEFAULT
We continue to be supportive of taking default risk in investment grade and high yield markets, given the generally healthy state of corporate balance sheets. The levered default-risk (long first-loss tranches) trades have become very popular in the hedge fund and banking community. The good news is that most of the positions are hedged to some degree, either for spreads moves (via short mezzanine or index protection) or for defaults in riskier credits, or both. But the risk that most investors have in this trade is a quick jump to default from an unexpected name, or a sudden reshaping of the risk distribution for an entire sector (like we saw in the insurance space, but on a bigger scale). At least based on the few examples of this in history, this risk is largely random, as it has involved events that were difficult to predict, including fraud and litigation. So today’s generally healthy balance sheets do not necessarily reduce this risk factor, given its randomness, although the focus on corporate governance in the US since 2002 may help to mitigate it.
Please see additional important disclosures at the end of this report.
305
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 45
exhibit 1
The Impact of Stress – 0-3% CDX Tranche
Scenario Widest credit moves up 1000 bps
Unhedged -6.3%
Delta Hedged 0.4% -2.1%
Tightest credit moves up 1000 bps
-6.6%
10 Wide credits up 100 bps
-6.0%
1.0%
10 Tight credits up 100 bps
-5.7%
0.7%
All credits move 8 bps each
-5.3%
0.8%
Widest credit defaults with 40% recovery
-12.8%
-1.0%
Tightest credit defaults with 40% recovery
-14.9%
-8.5%
Widest credit defaults with 0% recovery
-22.4%
-2.7%
Tightest credit defaults with 0% recovery
-24.2%
-13.5%
Source: Morgan Stanley
To demonstrate the jump risk, we calculated model-based losses that can occur in both equity and mezzanine tranches of the 125-name CDX index (see Exhibits 1 and 2). A tight or wide trading name moving 1,000 bp wider would result in a loss in excess of 6% for a 0-3% tranche if the credit were unhedged. If the credit were delta-hedged, there would be a slight gain if it were the widest trading name in the portfolio (because more protection would have been purchased) but still a significant loss (of 2.1%) if it were a tight trading name. If the jump risk were more sector related (10 names moving 100 bp wider each), the losses would still be significant in the unhedged case, but slightly less than in the one credit case. If jumping directly to default, the losses get quite a bit bigger for the equity tranches, as one would expect. At 40% recovery, the losses are 12.8% and 14.9%, respectively, for the widest and tightest names. The tightest name defaulting results in a higher loss because the tail of the portfolio does not shrink, as it would if the widest name defaulted. Even in the hedge case, losses occur because the deltas would not be precise enough for the jump risk. A more extreme case occurs with a 0% recovery, where a delta-hedged position on a tight trading name would result in a 13.5% loss on the 0-3% tranche, quite significant for a name that was delta hedged. If it were the widest name, the loss would have been only 2.7%.
306
exhibit 2
The Impact of Stress – 3-7% CDX Tranche
Scenario Widest credit moves up 1000 bps
Unhedged -1.3%
Delta Hedged 0.5% 0.1%
Tightest credit moves up 1000 bps
-1.9%
10 Wide credits up 100 bps
-2.8%
0.1%
10 Tight credits up 100 bps
-2.8%
0.0%
All credits move 8 bps each
-2.7%
-0.1%
Widest credit defaults with 40% recovery
-1.5%
1.6%
Tightest credit defaults with 40% recovery
-2.5%
0.4%
Widest credit defaults with 0% recovery
-3.5%
1.8%
Tightest credit defaults with 0% recovery
-4.6%
0.3%
Source: Morgan Stanley
For a junior mezzanine tranche (3-7%), the decision to delta hedge individual credits does protect the tranche from relatively large losses. However, delta hedging specific credits is very uncommon among mezzanine tranche investors, unless it is part of a larger correlation book. As such, losses as much as 4.6% can occur from a single default on a tight trading name (with 0% recovery) despite the fact that there is 3% subordination in the tranche. CREDIT CYCLE EXPERIENCE
A common criticism and fear inherent in the structured credit markets revolves around the relative lack of market experience among players in this “youthful” business. Although the majority of today’s users of correlation instruments have entered this market over the past 18 months, there is still a sizeable amount of users on both the dealer and investing side of the business who lived through the 2001-2002 credit cycle in seats where they had to manage structured credit risk, which we do find comforting. Yet, the experience gap in the marketplace is a risk, and will cause some dislocation when idiosyncratic risk rises. BIG ENOUGH TO BLAME
While it has become quite common in the credit markets to blame structured credit activity on the rapid tightening and spread compression that has occurred, it should be noted that for anyone comfortable with today’s spread levels, a disappearing structured credit bid can have an equal but opposite impact on the market. Will this draw even more cynicism from the larger credit community? The answer is probably yes. We do not see the structured credit bid fading away in the near-term, but over the medium-term the risk is certainly there. The key message is that there is dislocation risk in the credit markets that can be caused by “unexpected” flows. An unforeseen unwind from a hedge fund is another example, which could be related to issues at the fund, or even a flow triggered by the increasingly large community of fund of funds. We should note, though, that quite a few investors are waiting on the sidelines for an attractive invitation to get involved in this space, and such a flow-related dislocation might serve to be their calling.
Please see additional important disclosures at the end of this report.
307
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 45
UPFRONT PAYMENTS ARE PRE-FUNDED LOSSES
While the unfunded nature of most of the activity in the correlation space can lead to quite a bit of leverage deployment, we are comforted by the fact that the riskiest tranches require sizeable upfront payments to the seller of protection. The upfront payments are typically “posted” back to the counterparty, and effectively serve as a collateralization of the default exposure inherent in the first loss positions (see Exhibit 3). exhibit 3
Pre-Funded Losses – Upfront Coverage from Defaults
14% 12% 10% 8% 6% 4% 2% 0% 0-3% IG 5 Year
0-3% IG 10 Year
0-10% HY
10-15% HY
Source: Morgan Stanley
In the case of investment grade CDX, the 5-year upfront is sufficient to cover the first 1% of losses in the index (2 defaults with 40% recovery) while for the 10-year index, the upfront can cover roughly the first 1.4% of losses (3 defaults at 40% recovery). In high yield, today’s typical upfront on the first-loss tranche covers the first 7.6% losses in the 100-name index (12 defaults at 35% recovery), while the 10-15% tranche’s upfront covers 12% of losses (18 defaults at 35% recovery). MODELS, TRANSPARENCY AND THE POWER OF BLOOMBERG
Ironically, skepticism is increasing at a time when structured credit is more transparent than it has ever been. The transparency is actually demonstrating the option-like behavior of many of the instruments, which is a dramatic improvement from years ago when these risks were less well understood. We find this comforting. However, ratings have always been and will continue to be important guidelines for investors, but ratings are based purely on expectations of default, not on pricing movements, so pricing shocks can exist, which is a concern. So are the differences between the various agencies, especially when they are exploited. Market standard correlation models have existed for the past nearly two years, and today the introduction of a correlation model on Bloomberg will undoubtedly have a long-lasting impact on the market, much like the role Bloomberg functionality has played in many other corners of the fixed income markets. Yet, as we have discussed on numerous occasions, models will evolve over time, which can result in risk being repriced, like the move to a base correlation standard did earlier this year. We do worry about investors who have been too “married to the models” in their approach.
308
ACCOUNTING IMBALANCES
As we alluded to earlier, one of the imbalances in the market that concerns us is related to differences in accounting practices within the investment community. A good portion of the sellers of mezzanine and senior protection in the correlation space are investors who can carry the risk on an accrual basis, thereby mitigating much of the mark-to-market risk. In a moderately negative credit environment, they are unlikely to be forced sellers. So why are we concerned? We are concerned partially because those who have bought protection on the same tranches are generally mark-to-market investors, and may be forced to unwind this protection (as a hedge against something else that is falling in price) before its full hedging potential is realized. CORRELATION CYNICISM
Skepticism will always exist when the structure of markets change. We are by no means preaching that all is perfect in this new world, but we feel confident in challenging some of the correlation cynicism, given the progress that has been made in the structured credit space. But we do expect higher volatility, market dislocations, and a bit of frustration going forward, as investors manage credit portfolios along this new landscape.
Please see additional important disclosures at the end of this report.
309
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 46
Balancing Liquidity and Creativity
January 28, 2005
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Ajit Kumar, CFA The index tranche market continues to dominate liquidity, flows, and investor participation within the structured credit space. In fact, many newer investors associate the term ‘structured credit’ directly with the index tranches, even though the market is much broader than that. We do not deny that the hyper-growth in popularity of index tranches has opened up the market to many new investors and created much needed transparency and standardization, not to mention important valuation benchmarks and a two-way market. Clearly, there is more liquidity today in the index tranches that anyone could have ever imagined. Yet, the ease with which one can use index tranches has forced many investors into the same trades, both thematically and credit wise. For investors with reasonable amounts of structured credit exposure, but little diversity in terms of underlying portfolios and instruments, the resulting concentration risk can be harmful, triggered by both technical themes associated with activity in the liquid instruments as well as significant credit-specific events. Both of these triggers are very relevant in today’s market. From an opportunity set perspective, the structured credit market remains largely a bespoke one, where creativity, and the effort it requires, are the only barriers to entry. We encourage structured credit investors to keep this thought in mind as they hunt for value, as less trafficked routes may in fact be more scenic. THE VALUE OF LIQUIDITY
An age-old battle in the fixed income markets centers on the value of liquidity. Unlike in the equity markets, investors have many instruments to choose from, and one must constantly judge the importance of liquidity on the margin, given pricing and market opportunities. There is a tendency to seek out liquid opportunities, even though most investors will not require the liquidity most of the time. There have been studies done in the interest rate markets, for example, which show that off-the-run instruments outperform on-the-runs over long periods of time. CREATIVITY HAS ITS BENEFITS, EVEN IF IT’S MORE WORK
Getting back closer to home, our main point here is to encourage credit investors generally and structured credit users in particular to re-consider the importance of liquidity on the margin, as their exposures grow in absolute size. There can be some price advantages for seeking out less liquid opportunities, but more importantly, the diversity in credit exposure and risk profile is a key consideration.
310
With respect to balancing liquidity and creativity, there are three points we would like to make. First, the argument that a pure liquidity focus results in lots of credit overlap is an easy one to make. If every long, short, or long/short combination is exposed to largely the same sets of credits, then bolts from the blue will have an even more magnified impact. exhibit 1
Not Much in Common – Spread Distribution of CDX and a Seasoned Synthetic CDO
% of Portfolio 35% IG CDX 3
30%
Early Synthetic CDO 25% 20% 15% 10% 5% 0% 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 >200
Spread (bps)
Note: Synthetic CDO used for descriptive purposes is EPOCH 2000-1, issued 12/2000 with 3 years left to maturity from today. Source: Morgan Stanley
Second, the shapes of credit distributions can be vastly different. The index, which is frequently rebalanced and has a generally compressed distribution today, may look very different from a seasoned synthetic CDO tranche that was constructed during a different part of the credit cycle (see Exhibit 1). Given the distribution shape differences, tranches on these two portfolios would likely have fairly different risk and return characteristics as they may be driven more by credits in the portfolio than market-wide moves (i.e., they could be relatively uncorrelated, no pun intended). Third, we live in a very technical market today, and the technicals associated with CDX (because it is a traded instrument) have been somewhat frustrating for anyone who uses indices as a reflection of market activity. We argue that the most liquid tranches can also be very technical in nature. As such, we encourage investors to have some diversity in the exposure they have to any one technical theme.
Please see additional important disclosures at the end of this report.
311
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 46
THE IMPACT OF SINGLE-NAME EXPOSURE – INVESTMENT GRADE
We take forward the first point, as it has a tremendous amount of relevance today. Despite all of the spread compression in the market, there is still a decent number of story credits in the investment-grade and cross-over space, from General Motors and Bombardier to the M&A risk associated with both wide and tight trading names. In fact, given where we are in the cycle and the generally healthy state of corporate balance sheets, it is this credit-specific risk that is likely the biggest issue for structured credit investors today, rather than big swings in market-wide average spreads. We highlight the P&L associated with some severe credit-specific events (involving one or two credits at the most) in Exhibit 2 for the standard investment grade index tranches (protection seller’s perspective). We highlight a few points for anyone with large amounts of unhedged exposure to market standard portfolios. •
If the widest trading name in CDX tightened significantly (to 100 bp) the equity and 3-7% 5-year tranches would outperform the index tremendously, while performance of the other tranches would be more in line with the index. In the 10year tranches, the outperformance would extend to the 7-10% tranche as well.
•
A tight name becoming distressed or defaulting immediately would result in a 12% to 15% loss for the investment grade 5 year equity tranche but a lower 7% to 10% loss for the 10-year 0-3% tranche. This is related to the much higher upfront payment for the latter, which also reduces the tranche’s sensitivity to spread changes. The opposite is true for the mezzanine tranches, where the impacts are much worse for 10-year versus 5-year.
•
Two names defaulting, however unlikely that may seem in investment grade today, can cause P&L outcomes many times that of the index in equity and mezzanine tranches, especially for 10-year tranches.
The key message is that most large investment grade credit portfolios tend to contain exposure to more than 125 names (the number in CDX) and also tend to be debtmarket capitalization-weighted instead of being equally-weighted. While intentional single-name concentration is certainly a subordinate tranche investment strategy, for those who have been forced into this state because it is easy, we encourage some credit selection exercises.
312
Please see additional important disclosures at the end of this report.
313
Source: Morgan Stanley
Tranche P&L 0-3% 3-7% 7-10% 10-15% 15-30% Index Tranche P&L (Delta Hedged) 0-3% 3-7% 7-10% 10-15% 15-30%
Tranche P&L 0-3% 3-7% 7-10% 10-15% 15-30% Index Tranche P&L (Delta Hedged) 0-3% 3-7% 7-10% 10-15% 15-30%
-7.1% -8.5% -3.6% -1.6% -0.4% -0.7% -2.4% -2.8% -0.4% 0.1% 0.3%
0.6% 0.6% 0.1% 0.0% -0.1%
-3.3% -0.1% 0.6% 0.3% 0.1%
0.7% 0.0% -0.1% -0.1% 0.0%
1.7% 2.0% 0.9% 0.4% 0.1% 0.2%
-12.6% -4.1% -1.1% -0.4% -0.1% -0.6%
2.7% 0.8% 0.2% 0.1% 0.0% 0.1%
-6.4% -1.2% 0.3% 0.4% 0.3%
-9.8% -5.3% -1.9% -0.8% -0.2% -0.5%
-7.0% 0.9% 0.9% 0.4% 0.2% IG 10 Yr Series 3 CDX
-15.0% -2.4% -0.5% -0.2% 0.0% -0.5%
-11.9% -2.8% 0.5% 0.8% 0.6%
-18.6% -10.9% -4.0% -1.6% -0.4% -1.0%
-12.9% 1.3% 1.6% 0.8% 0.3%
-28.9% -5.5% -1.2% -0.4% -0.1% -1.0%
-6.3% -0.5% 0.5% 0.5% 0.3%
-8.1% -2.7% -0.7% -0.2% 0.0% -0.3%
-6.6% 1.1% 0.8% 0.4% 0.1%
-12.1% -1.2% -0.2% 0.0% 0.0% -0.3%
Widest Credit Defaults with 40% Recovery
-12.4% -1.7% 1.0% 0.9% 0.5%
-16.7% -6.9% -1.9% -0.6% -0.1% -0.7%
-13.2% 1.9% 1.7% 0.8% 0.3%
-25.5% -3.3% -0.5% -0.1% 0.0% -0.8%
2 Widest Credits Default with 40% Recovery
Investment Grade Tranches – The Impact of Severe Single-Name Events (Seller of Protection Perspective)
Widest Credit Tightest Credit Tightest Credit Defaults 2 Tightest Credits Default Tightens Significantly Widens 1000 bps with 40% Recovery with 40% Recovery IG 5 Yr Series 3 CDX
exhibit 2
314
Source: Morgan Stanley
Tranche P&L 0-10% 10-15% 15-25% 25-35% 35-100% Index Tranche P&L (Delta Hedged) 0-10% 10-15% 15-25% 25-35% 35-100% -0.9% -1.5% -0.9% -0.4% -0.1% -0.4% -0.3% -0.3% -0.1% 0.0% 0.1%
2.2% 1.8% 0.8% 0.2% 0.0% 0.5% 1.3% 0.3% -0.3% -0.3% -0.1%
-1.6% 0.0% 0.6% 0.4% 0.2%
-2.9% -2.5% -1.3% -0.4% -0.1% -0.8%
-5.1% -0.9% 1.8% 1.4% 0.7%
-10.5% -10.5% -5.5% -1.8% -0.3% -3.0%
-0.4% 0.4% 0.3% 0.2% 0.0%
-0.7% 0.0% 0.0% 0.0% 0.0% -0.1%
Widest Credit Defaults with 35% Recovery
-2.4% 1.9% 1.7% 0.8% 0.3%
-3.7% -0.6% -0.2% 0.0% 0.0% -0.8%
4 Widest Credits Default with 35% Recovery
High Yield Tranches – The Impact of Severe Single-Name Events (Seller of Protection Perspective)
chapter 46
Widest Credit Tightest Credit Tightest Credit Defaults 4 Tightest Credits Default Tightens Significantly Widens 1000 bps with 35% Recovery with 35% Recovery HY 5 Yr Series 3 CDX
exhibit 3
Structured Credit Insights – Instruments, Valuation and Strategies
THE IMPACT OF SINGLE-NAME EXPOSURE – HIGH YIELD
The high yield situation is somewhat different though, given how the standard tranches are structured, the smaller number of names that trade in the market, and the limited number of off-the-run synthetic structured credit opportunities available (although there are plenty of seasoned cash CBOs and CLOs, see Chapter 72). From a sensitivity perspective, since many more defaults are priced in based on the industry standard risk-neutral models, the impacts are less extreme on relative basis. For example, if the HY CDX index experienced 4 near-term defaults, the models show that the 0-10% tranche losses would vary from 3.7% to 10.5%, while the 10-15% tranche (which is equity-tranche-like in risk) would vary from 0.6% to 10.5% (see Exhibit 3, from a protection seller’s perspective). With 4 near-term defaults on wide trading names, based on model valuations and no change in correlation, the 10-15% tranche would experience positive moves on a delta neutral basis, although we caution that 4 quick defaults would likely cause spread widening and adverse correlation moves that could more than offset these model based figures. CREDIT SELECTION
In case the numbers cloud the key points, we reiterate that in today’s market environment, much of the risk in structured credit strategies will be driven by credit selection and market technicals, and less so by the generic scenarios like the market moving 50 bp wider or tighter. For investors who have been forced into the credit exposure by liquidity, we encourage balancing liquidity with credit selection creativity.
Please see additional important disclosures at the end of this report.
315
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 47
Correlation and Credit Cycles
February 4, 2005
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Ajit Kumar, CFA As we progress deeper into this part of the credit cycle, valuations continue to compress for the most part, and investors are getting comfortable with taking on more, and in many cases different kinds of credit risks, particularly in the structured credit space. We have devoted quite a bit of our research efforts over the past several months to the high yield opportunity in structured credit, given the secular changes that are occurring in the high yield space with respect to derivatives, juxtaposed with the prominence of the cash CDO market. We continue to view this as an interesting opportunity, especially for fundamentally minded investors, but we also see opportunity across the term structure of tranches. Within the investment grade structured credit space, a change that is likely more cyclical than secular in nature is the addition of liquidity and investor interest into maturities beyond 5 years. There is quite a bit of bespoke deal flow in longer maturities today, particularly in 7 years, in part because of the steepness of credit curves. Many long-term credit investors are establishing structured credit positions in longer-dated credit, particularly at the mezzanine and senior levels. However, since the bespoke market is somewhat fragmented, the most transparent part of the market beyond five years is the market for standardized 10-year index tranches, which today is perhaps seeing more flows than the benchmark 5-year tranches. Yet, we caution that the 10year index tranche market appears very technical, with flows dominated by the dealer community (hedging risk coming from the bespoke market) and hedge funds, which is not unlike the early days of 5-year index tranche trading. Given this newfound activity in longer maturities, we provide our readers with a few thoughts on how to compare the various opportunities. Although they are optically similar, the risk and return profiles of 5- and 10-year index tranches are quite different, with the 10-year standard tranches being more similar to the high yield tranches, calibrated for expected loss differences. The steep credit curve dominates relative valuation, and for anyone with strong credit cyclical views that disagree with curve steepness, the 10-year liquidity point provides ways of implementing these views.
316
A STEEPER CREDIT CURVE
One of the key motivators for taking longer-dated credit risk in structured form has been a fairly sharp steepening of the 5s-10s credit curve over the past 15 months, in an otherwise tightening market environment. When we first argued that the credit curve should be steep (see “Getting Short the Long End,” October 10, 2003), the 5-year premium on the index was 67 bp, and there was approximately 10 bp of steepness from 5 to10 years (a 15% premium to 5-year risk). Today, the 5s-10s curve is 25 bp steep, on a much lower base (45 bp for 5 years), which makes 10-year risk trade at a 56% premium to 5-year risk. exhibit 1
CDX 5s-10s – Curve Steepens While Spreads Tighten
80
30
70 25 60 20 50 40
15
30 5 Year (Left Axis) 20
10
5s-10s Curve (Right Axis) 5
10 0 Oct-03
Jan-04
Apr-04
Jul-04
Oct-04
0 Feb-05
Source: Morgan Stanley
What is the right level of steepness for credit curves, given where we are in the economic cycle? History tells us that, for 10-year investment grade credits, there is generally more default risk in the back 5-years than in the front 5-years (50% more on average, see Exhibit 2 for BBBs). Yet, previous studies suggest that the difference can be even greater if we are in the early part of the credit cycle recovery, although we would argue that at this point the credit recovery cycle is more mature than that (see “Getting Short the Long End,” October 10, 2003, for our early cycle views).
Please see additional important disclosures at the end of this report.
317
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 47
exhibit 2
BBB Default Experience, First 5 Years vs. Back 5 Years
8.0% Current 0-5 Year 7.0%
Shifted 5-10 Year
6.0% 5.0% 4.0% 3.0% 2.0% 1.0%
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
0.0%
Source: Morgan Stanley, Moody’s
TIME, TRANCHES AND CREDIT CYCLES
Today’s steep curves indicate significant increases in credit risk 5 years forward. This difference between 5- and 10-year risk has dramatic impacts on risk-neutral correlation models used for pricing tranched risk, and can result in relative value opportunities for those willing to trade credit cycle views against the models. In Exhibit 3, we illustrate the proportion of 10-year expected losses priced into the first 5 years for each tranche of CDX, both today and in August 2004, when 5-year spreads were wider (at 66 bp) and curves flatter. exhibit 3
How Much of 10-Year Risk is in the First 5 Years?
90% CDX 8/9/2004 (66/87)
80%
CDX Today (47/73)
70% 60% 50% 40% 30% 20% 10% 0% 0-3%
3-7%
7-10%
10-15%
15-30%
Source: Morgan Stanley
The results are intuitive; the majority of risk for the 0-3% tranche is associated with near-term default risk, while the majority of losses for the other tranches are associated with back-end risk. The 0-3% tranche is already in the money (or underwater, see below) for 5-year risk at today’s spread levels, and therefore the incremental risk in the 5-10 year period is relatively small. At the same time the allocation of risk for the more senior tranches is much more stable and biased toward the back 5 years because these tranches remain out of the money (above water) for both 5- and 10-year risk.
318
Today’s tight spread environment results in an increased proportion of the risk in the 37% tranche being priced in the back five years when compared to what we saw in August. For those looking to play the timing of default risk views, this part of the capital structure offers opportunity because it prices like out-of-the-money protection for 5-year risk and in-the-money protection for 10-year risk. COMPARING CAPITAL STRUCTURES – HOW FAR ABOVE WATER?
With a plethora of liquid index tranches that trade in the market today (6 families between the US and Europe), there can be quite a bit of confusion for anyone who simply attempts to compare them. We find a simple, yet still useful method is to think of the tranches as being above or underwater, from the perspective of portfolio expected losses (which are implied from spreads). For example, with 5-year CDX in the US at 45 bp, the expected losses over 5 years are approximately 2.2% and the most subordinate tranche has a 3% detachment point. Our simple ‘below or above water’ metric is the ratio of the two or 140% (see Exhibit 4). The standard equity tranches for the 10-yr CDX (0-3%) and the HY CDX (0-10%) are completely underwater, and have scores below 100%. A simple glance of the 3 tranche families in the US, based on this metric, shows how different 5- and 10-year tranches are from each other, with the 10-year IG and 5-year HY having surprising similarities. To be clear, this is a function of how the market chose to standardize tranches, not a statement about risk similarities between investment grade and high yield. exhibit 4
How Far Above Water – Tranches vs. Expected Losses
800% 5Y CDX 700% 10Y CDX 600%
5Y HY CDX
500% 400% 300% 200% 100% 0% Tranche1: 0-3% 0-10%
Tranche2: 3-7% 10-15%
Tranche3: 7-10% 15-25%
Tranche4: 10-15% 25-35%
Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
319
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 47
CORRELATION REGIMES
While, from an expected loss perspective, the standard 10-year investment grade CDX tranches resemble their HY CDX cousins, from an implied correlation perspective, the 5-year and 10-year investment grade tranches are much more in line (see Exhibit 5). We have discussed, in previous research, a variety of reasons why correlation means something very different in the high yield space (see Chapter 16). Under some reasonable assumptions, spreads in the high yield index imply roughly 12.5% losses over 5 years. Based on Moody’s data, we approximate historical 5 year loss experience for a comparably rated index has ranged from 6% to 23%. Meanwhile, even for the 10-year maturity in investment grade, we estimate the index losses have ranged from 0.4% to 3.0%. While a result marginally worse than expectations in high yield could easily generate losses in excess of the 10-15% tranche detachment point, even the worst case scenario (historically) in investment grade barely touches the 10year 3-7% tranche (assuming a market portfolio), despite the fact that these tranches are reasonably comparable based on the analysis in Exhibit 4. This exposure to default rate volatility in likely one reason why the 10-15% tranche in high yield prices cheap (based on correlation skew) relative to the closest investment grade tranches. exhibit 5
70%
Implied Base Correlation
60%
5Y CDX
50%
10Y CDX 5Y HY CDX
40% 30% 20% 10% 0% Tranche1: 0-3% 0-10%
Tranche2: 3-7% 10-15%
Tranche3: 7-10% 15-25%
Tranche4: 10-15% 25-35%
Source: Morgan Stanley
Additionally, expected losses do not capture the jump-to-default risk, as the models give this a very low probability of occurring. For such an event, the 5- and 10-year CDX 0-3% tranches have roughly the same absolute default exposure (27% of their notional), while the 0-10% HY tranche has only 10% of its notional exposed to a sudden jump to default. This is an argument for investment grade equity tranches trading cheaper on a correlation basis (lower correlation).
320
CORRELATION AND CREDIT CYCLES
We have touched upon curve ideas within the tranche space very early on in previous research (see Chapter 26). Yet, with a rather large liquidity injection in the 10-year part of the index tranche market today, and quite a bit of activity in longer-dated bespoke portfolio tranches as well, investors have much more choice in implementing structured credit ideas along the curve. Given where we are in the credit cycle, we continue to like taking shorter maturity default exposure through subordinate tranches, and hedging some of market risk of those positions with longer-dated tranche protection. For long-only strategies, one cannot ignore the steepness of credit curves, but understanding which tranches will behave like in- or out-of-the-money options across various maturities and asset classes is an invaluable addition to the structured credit toolbox.
Please see additional important disclosures at the end of this report.
321
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 48
How Big Is the Structured Credit Market? 2005 Update Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Vishwanath Tirupattur Primary Analyst: Pinar Onur Understanding the size and the scope of the bespoke market offers useful insight on the nature of the structured credit risk, leverage levels, and the market trends. To this end, we are providing our third study on the size of the bespoke market (see our original report, “How Big Is the Structured Credit Market?” April 8, 2005 and “How Big Is the Structured Credit Market? Summer Update” August 5, 2005). As we did in our previous two studies, we use issuance data from Creditflux, a UK provider of news and analysis on the credit derivatives and structured credit markets. We consider Creditflux the best source for bespoke structured credit activity in the market, although even Creditflux admits that it may only track 60% of the market. Our key takeaways from this study are as follows:
322
•
2005 was a busy year for the bespoke market – the total issuance almost doubled (1.9x) to $330 billion in 2005 from $116 billion in 2004.
•
$330 billion total issuance translates into $615 billion of credit risk on a delta adjusted basis ($329 BN in 2004). While the total delta adjusted credit risk number is significantly higher in 2005 compared to 2004, the increase is lower in proportion to the increase in the total bespoke issuance (87% vs. 185%) due to the recent popularity of the super senior and senior tranches.
•
The issuance activity in the bespoke market picked up significantly in the second half of 2005 – 2H 2005 issuance stood at $221 billion, double the amount of issuance in the first half of the year ($109 billion).
•
In 2005, there has been an outflow from the mezzanine tranches to the other parts of the capital structure both on a total issuance and delta adjusted risk basis.
•
CDO-squared issuance dropped from 14% to 5% of notional, which is a mezzanine phenomenon, as well.
•
Synthetic structured finance CDO issues grew slightly in 2005 ($49 billion total notional versus $48 billion).
•
Unfunded structures continue to dominate the market. The funded structures constituted only 22% of the market in 2005 – down from 33% in 2004. Furthermore, the split between funded and unfunded structures is fairly even for USD denominated issuance while EUR denominated issuance is dominated by unfunded structures (75%). A higher participation in the bespoke market by the US insurers/money managers might explain this last phenomenon.
Please see additional important disclosures at the end of this report.
323
257,846 71,476
Unfunded Funded 77,744 38,032
Source: Morgan Stanley
2005
Vintage 2004
Notional
329,322
Total 115,776 49%
USD 45% 44%
EUR 49% 31%
Managed 37%
69%
Static 63%
Portfolio Style
34%
28%
38%
Credit Risk (Delta Adjusted) Senior + Equity Mezzanine Super Senior 25% 51% 25%
Structured Credit Market Summary – Bespoke Issuance ($ MM)
Specified Currency of Notional
exhibit 1
615,188
Total 329,063
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 48
THE STRUCTURED CREDIT BID LIVES ON
Despite significant credit volatility in 2005, bespoke activity almost doubled in size since 2004. Meanwhile, the delta adjusted credit risk grew less at 87% due to the shift out of mezzanine tranches into more senior parts of the capital structure, which we discuss below in detail. The follow-on impacts of this risk shift support a decline in CDO-squared transactions and alter the risk profile of outstanding bespoke structures, to be less idiosyncratic and more systematic or even systemic in nature. On the synthetic SF CDO front, the expected strong growth did not materialize in 2005. We attribute this phenomenon partly to a wait-and see approach on the part of the investors. OUT OF MEZZANINE, INTO EQUITY AND SENIOR
Significant spread tightening in the mezzanine tranches (about 200 bp for 3-7% 5-year) and the recent changes in the S&P CDO model have slowed down mezzanine issuance and resulted in a shift of risk from mezzanine into equity, senior and super senior tranches. On a delta-adjusted basis, bespoke mezzanine tranches represent about $173 billion of credit risk in 2005 (28% of the total) versus $167 billion last year (51% of the total). exhibit 2
Delta-Adjusted Credit Risk
Equity
Junior Mezz
Senior Mezz
2004
4,136
5,188
16,389
27,325
62,738
115,776
2005
10,518
7,607
13,901
70,332
226,964
329,322
Avg. Attachment
0.0%
2.5%
4.5%
9.1%
16.0%
Avg. Detachment
2.8%
3.7%
6.0%
12.9%
83.8%
Estimated Delta
20.0
10.0
7.0
2.0
0.4
Senior
Super Senior
Grand Total
Tranche Notional
Delta Adj. Risk 2004
82,713
51,881
114,724
54,651
25,095
329,063
2005
210,364
76,070
97,304
140,665
90,785
615,188
Source: Morgan Stanley, Creditflux
The growth in super senior activity has been the most impressive (262%), even on a delta adjusted basis. The increase in equity tranche (154%) and senior tranche (157%) related bespoke activity has also been dramatic. The bespoke issuance has been sustained even in the mezzanine tranches ($22 billion in both 2004 and 2005) which might be a result of the increased activity in the 7-year part of the curve; especially during the second half of 2005 (we draw this conclusion anecdotally as the data does not contain any information on the maturity). This increase
324
in 7-year mezzanine type tranches has been supported by steep credit curves, a positive basis, and newfound activity in 7-year CDX tranches (see “Checking the Pulse of the Market, Post-Volatility,” July 15, 2005, and Chapter 18). Average attachment and detachment point for senior and super senior tranches are slightly up (1-3%) since our studies in April and August. On the other hand, the average attachment/detachment points have not changed significantly for mezzanine and equity tranches. exhibit 3
2004 – Year of Mezzanine
60% 54%
Tranche Notional/Total Credit Risk/Total
50%
40% 35%
30% 25%
24%
20%
16%
17% 14%
10%
8% 4%
4%
0% Equity
Junior Mezz
Mezz
Senior
Super Senior
Source: Morgan Stanley, Creditflux
exhibit 4
2005 – Out of Mezzanine, into Equity and Senior
80% 69%
Tranche Notional/Total
70%
Credit Risk/Total
60% 50% 40%
34%
30% 21%
20%
23%
16%
15%
12%
10% 3%
2%
4%
0% Equity
Junior Mezz
Mezz
Senior
Super Senior
Source: Morgan Stanley, Creditflux
Please see additional important disclosures at the end of this report.
325
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 48
While issuance of funded tranches is only a small piece of the overall structured credit markets, they provide good information both on the nature and amount of credit risk being taken out of the system (the assumption being that funded tranches are unlikely to be sold back quickly, given the costs associated with creating credit-linked notes). In both 2004 and 2005, rated tranches represent 55-60% of all funded notes, but AAAs represent 45% this year versus only 30% last year, demonstrating, once again, the move toward senior risk. The corresponding drop has been in AA and A rated tranches. exhibit 5
2004
Funded Structured Credit – More Activity in AAAs in 2005
A 8% AA 13% NR 44%
AAA 30% BBB 4%
HY 1%
20 0 5 A 4%
AA 6%
NR 40%
AAA 45% BBB 3%
HY 2%
Source: Morgan Stanley, Creditflux
326
LOOKING FORWARD
Majority of the bespoke issuance activity took place in the second half both in 2004 and 2005 (61% and 67% respectively). We believe that 5-year mezzanine type issuance boosted the 2H 2004 activity, while the senior/super senior trade and 7-year mezzanine type issuance were in the driver seat in the second half of 2005. With the mezzanine spreads sailing along all time tights, especially in the 5 and 7 year space, it will be interesting to see where the issuance activity will shift to. In the mezzanine space, the activity might move towards 10-year mezzanine type risk which would shift the risk structure in the market as these tranches behave more like equity. The risk might also continue to flow into equity and senior tranches. The amount of delta adjusted credit risk in the system grew 87% in 2005, bolstered by the rapid growth in the bespoke market. A relatively broad credit stress - such as multiple defaults in the auto sector including GM – would have implications for the mezzanine tranches and might cause a slow down in the overall bespoke market. Another trend to follow will be the development of the synthetic SF CDO market. We continue to believe that the activity in this space will pick up as ISDA standards for structured finance CDS and pay-as-you-go structures become more widely accepted. Managed structures constituted ~35% of the market in 2004 and 2005. Preliminary 2006 data will tell whether the growing popularity of self-managed structures tip the scale in favor of managed structures. The self-managed structures are attractive for investors who are looking to leverage their own credit-picking skills as well as those investors who seek ratings and income stability (see Chapter 41). Finally, the banks, insurers and reinsurers are likely to increase their level of involvement in the structured credit market as a result of upcoming changes in FASB, Basel II and IAS. The impact these players will have on the market will be another trend to follow, particularly on self managed and funded structures.
Please see additional important disclosures at the end of this report.
327
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 49
A Loan in the Storm
May 13, 2005
Primary Analyst: Vishwanath Tirupattur Primary Analyst: Sivan Mahadevan Corporate credit markets have continued to experience a high degree of volatility through April and early May. Standard and Poor’s downgrade of GM and Ford came sooner than anticipated by many credit investors. The degree of idiosyncratic volatility unleashed by the storm in the auto and auto-parts sectors has not been seen in corporate credit since 2002. The turmoil in the standardized credit derivatives index tranches experienced in the last few days (in this context, see Chapter 31 for a discussion and analysis of recent developments) has resulted in a significant re-pricing of credit risk in index tranches. These developments, together with the insipid performance of global equity markets year to date and apprehension about prospects for economic growth have spurred investors to reexamine risk exposures, reallocate across asset classes and rebalance portfolios. Terms such as “flight-to-quality,” “contagion,” etc. are back in vogue. Given this background, we think it is an opportune time to revisit the performance of the leveraged loans sector relative to unsecured corporate debt, as well as CLO debt tranches relative to CBO debt tranches, particularly during periods of stress. We hope that investors will find this exercise useful as they consider the risk-return profile for different asset classes.
328
SECURED AND UNSECURED CORPORATE CREDIT: A COMPARISON
During the turmoil of the past few weeks, the secured credit market of leveraged loans has held up significantly better than the unsecured bond market. According to S&P LCD data, the weighted average spread of all outstanding loans was at 266 bp by April end, tighter by about 4 bp from the end of March. BB loan spreads were tighter by 8 bp and B loan spreads were virtually unchanged (Exhibit 1). While new issue loan spreads in April widened from their March-end levels and have continued to increase in the first few days of May, the degree of widening has been modest. exhibit 1
Weighted Average Spreads of Outstanding Loans
400 All Loans BB Loans 360
B Loans
320
280
240
200 Jan-98
Mar-99
Jun-00
Aug-01
Nov-02
Feb-04
Apr-05
Source: Morgan Stanley, The Yield Book and S&P LCD
In contrast, benchmark five-year investment grade bond index and high yield index spreads over the same period have widened by almost 50% and 30%, respectively. In terms of total returns, for the first four months of the year, loans have outperformed investment grade bonds by over 90 bp and high yield bonds by over 150 bp. Undoubtedly, technical pressures in the loan market have been strong and have contributed to the outperformance. According to S&P, inflows into traditional loan accounts exceeded available new-issue supply during April – after subtracting the amounts taken by hedge funds, finance companies and other proprietary accounts – by $1.2 billion. In stark contrast, high yield bond mutual funds have continued to suffer outflows for 11 successive weeks, totaling $5.9 billion, or 7% of the total assets of the funds according to AMG Data Services.
Please see additional important disclosures at the end of this report.
329
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 49
While the current favorable technical picture may be a recent and possibly transient phenomenon, leveraged loan returns data over a longer timeframe reveal a picture of consistent excess returns performance with limited volatility. We use monthly data for the period January 1997 to April 2005, 1 to compare cumulative excess returns from leveraged loans with cumulative excess returns of benchmark indices of investment grade and high yield bonds (Exhibit 2). exhibit 2
Comparing Excess Returns
20%
IG Bonds HY Bonds
10%
Leveraged Loans
0%
-10%
-20%
-30%
-40% Jan-97
Jun-98
Oct-99
Mar-01
Jul-02
Dec-03
Apr-05
Source: Morgan Stanley, The Yield Book and S&P LCD
Over the period 1997 to date, the cumulative excess returns of leveraged loans are superior to both investment grade and high yield bonds. Returns have been consistent and with limited downside volatility.
1
Consistent data series on leveraged loan spreads we have available starts in January 1997 which dictated our time frame for this analysis.
330
In Exhibit 3, we compare summary statistics of excess returns data among leveraged loans and investment grade and high yield bonds. Using the Institute for Supply Management (ISM) Manufacturing Index as a proxy for general economic conditions, we define August 2000 to January 2002 as a period of sustained economic downturn when the index was below 50. 2 exhibit 3
Summary Statistics of Monthly Excess Returns (%) (January 1997 – April 2005) IG Bonds
HY Bonds
Leveraged Loans
Mean
0.04
-0.01
0.12
Std Dev
0.70
2.60
0.62
Sharpe Ratio
0.06
-0.003
0.20
0.14
-0.64
-0.06
Jan 1997 – Apr 2005
Aug 2000 – Jan 2002 Mean Std Dev
0.78
3.47
0.78
Sharpe Ratio
0.17
-0.18
-0.08
Source: Morgan Stanley, The Yield Book and S&P LCD
Over the longer timeframe, leveraged loans demonstrate a clear outperformance over both investment grade and high yield bonds. This is true both in terms of absolute returns and in the comparison of Sharpe ratios. Unsurprisingly, investment grade bonds outperformed both leveraged loans and high yield bonds on a risk-adjusted basis during the last bout of economic downturn. However, it is worth noting that leveraged loans significantly outperformed high yield bonds in both absolute and risk-adjusted terms.
2
The index was also below 50 for the seven months between June 1998 and December 1998. Because that represents very few data points, we are reluctant to draw significant conclusions from that period. However, the general tenor of the outperformance of the investment grade bonds over leveraged loans and high yield bonds and the outperformance of leveraged loans over high yield bonds was also true for that period.
Please see additional important disclosures at the end of this report.
331
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 49
CLO VS. CBO PERFORMANCE – ANOTHER LOOK
As indicated above, leveraged loans as the underlying sector have done significantly better than high yield bonds. We turn to the performance of their securitized counterparts in the form of CBOs and CLOs. In terms of ratings stability and performance, CLOs have done significantly better than CBOs, accounting only for 5.7% of all CDO downgrades. In contrast, high yield CBOs contributed to 61.1% of all CDO downgrades (see “Understanding Asymmetry in CDO Ratings Migration,” February 11, 2005). Unfortunately, we do not have the luxury of benchmark indices and their returns data series to facilitate this comparison in the same manner as the underlying collateral markets. We rely on another series of metrics to evaluate the performance of CLOs versus CBOs. Using aggregate data from S&P, we can observe time series of the overcollateralization (OC) spread as an indicator of performance for different vintages of CBOs as well as CLOs. The senior OC spread is the difference between the actual ratio and the required minimum ratio of the senior-most overcollateralization test, as a percentage of the required minimum ratio weighted by the size of each transaction’s collateral pool. Similarly, the subordinate OC spread is the difference between the actual ratio and the required minimum ratio of the subordinate-most overcollateralization test, as a percentage of the required minimum ratio weighted by the size of each transaction’s collateral pool. OC spreads should be positive at the outset of the transaction, meaning that the available collateral coverage to support the debt tranches is more than what is deemed sufficient. If experienced credit losses in the collateral pool over the life of the transaction are below expected credit losses, this spread narrows over time and translates into returns for the equity tranches. As transactions start to de-lever, OC spreads for healthy transactions could rise. When the spread turns negative, the actual collateral coverage is inadequate to support the debt tranches, which generally triggers the redirection of cash flows in the transaction to debt tranches (in order of seniority) and away from equity and other subordinated tranches. As such, a positive value for the OC spreads is a positive indicator of a transaction’s health and performance, and vice versa. We analyze data on the senior and subordinate OC spreads for different vintages of CLOs and CBOs. We focus on 2001 and prior vintages, which have a time series of performance long enough to make reasonable conclusions. These vintages also have the benefit of having gone through the vicissitudes of a full credit cycle.
332
In Exhibits 4 and 5 we compare the senior OC spread performance of CBOs and CLOs. The difference in the performance is striking. For CBOs, senior OC spreads started to decline for 1997-2000 vintages soon after the downturn in the economy and continued their downward trajectory into negative territory. 2001 vintage deals, structured during and soon after the economic downturn, have fared much better, maintaining positive and stable levels throughout the period. On the other hand, CLO senior OC spreads have remained positive and stable throughout the period, across all vintages, encompassing periods of economic downturn as well as expansion. The upward turns for the 1997 and 1998 vintages is a result of delevering. exhibit 4
CBO Senior OC Spread: 1997-2001 Vintages
40% 1997 Vintage 1999 Vintage 2001 Vintage
30% 20%
1998 Vintage 2000 Vintage
10% 0% -10% -20% -30% -40% Apr-00
Jan-01
Nov-01
Sep-02
Jun-03
Apr-04
Feb-05
Jun-03
Apr-04
Feb-05
Source: S&P
exhibit 5
CLO Senior OC Spread: 1997-2001 Vintages
35% 30% 25%
1997 Vintage
1998 Vintage
1999 Vintage
2000 Vintage
2001 Vintage
20% 15% 10% 5% 0% Apr-00
Jan-01
Nov-01
Sep-02
Source: S&P
Please see additional important disclosures at the end of this report.
333
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 49
Exhibits 6 and 7 depict a similar comparison for the subordinated OC spreads for CBOs and CLOs, respectively. Again, CLOs demonstrate a clearly superior performance relative to CBOs across all vintages. During periods of economic downturn, the subordinated OC spreads for CLOs declined for all vintages and even dipped into negative territory for the 1997 and 1998 vintages, but, with a pick-up in economic expansion, they changed their course and have been moving upwards since. The 1997 and 1998 vintages appear headed back into positive territory. In stark contrast, the decline in subordinated OC spreads for CBOs of all vintages has been sharp and has dipped significantly into negative territory, with the sole exception of the 2001 vintage. The return of the economy to an expansionary mode has not resulted in a significant improvement in the case of CBOs under this metric. In other words, subordinated tranches of CLOs recovered with the recovery in broader macro economic conditions and improvements in credit cycle but the subordinated tranches of CBOs did not. exhibit 6
CBO Subordinated OC Spread: 1997-2001 Vintages
20% 10% 0% -10% -20% -30% -40% -50% Apr-00
1997 Vintage
1998 Vintage
1999 Vintage
2000 Vintage
2001 Vintage
Jan-01
Nov-01
Sep-02
Jun-03
Apr-04
Feb-05
Source: S&P
exhibit 7
CLO Subordinated OC Spread: 1997-2001 Vintages
6%
1997 Vintage
1998 Vintage
1999 Vintage
2000 Vintage
2001 Vintage
4%
2%
0%
-2%
-4%
-6% Apr-00
Source: S&P
334
Jan-01
Nov-01
Sep-02
Jun-03
Apr-04
Feb-05
IMPLICATIONS GIVEN CURRENT MARKET CONDITIONS
For seasoned investors and followers of CDO markets, the differences we demonstrate between CLO and CBO performance should not be surprising. The point of revisiting the relative performance of CBOs and CLOs is to focus on lessons that can be drawn from historical performance, as we appear headed towards a downturn in credit cycle. It is no coincidence that leveraged loans and CLOs withstood downturns in the economy and credit cycle significantly better than high yield bonds. While we do not want to undermine the significance of structural protections and other structural differences between CBOs and CLOs, the performance of the underlying collateral is the predominant explanation for the differences in their performance. According to a recent study by Moody’s, 3 bond three-year default rates have been about 1.25 times greater than loan default rates, loss severity in the event of default has been about 2.25 times greater for bonds than for loans and overall loss rates have been about three times greater for bonds than loans. Exhibit 8 summarizes default, recovery and loss rates by rating category from the Moody’s study. Default, Recovery and Loss Rates by Rating Category
exhibit 8
Rating Baa
Default Rate 1.6%
Bonds Recovery Rate 37.4%
Loss Rate 1.0%
Default Rate 1.6%
Loans Recovery Rate 49.7%
Loss Rate 0.8%
Ba
5.3%
15.4%
4.5%
10.0%
69.6%
3.1%
B
21.1%
23.3%
16.2%
24.3%
70.3%
7.2%
Caa-C
51.7%
22.3%
40.1%
59.3%
66.0%
20.2%
Source: Moody's Investor Service
The superior performance of CLOs is really a result of relatively lower experienced defaults and higher recovery rates. As second lien loans and unsecured bond buckets increase in new issue CLOs, the risk of appearing more like CBOs will begin to rise, something investors should be both conscious and cautious of. While the jitters in the unsecured corporate credit and standardized index tranches have not affected leveraged loans or cash CDO markets broadly yet, we caution that technical winds can sweep into asset classes not so far affected. However, risk analyses and asset allocation exercises should be done in the context of risks and rewards being faced by all asset classes and prevailing economic conditions. We surely have concerns with respect to the CLO portfolio credit quality. Still, given overall economic uncertainty and turmoil in other competing asset classes, the resiliency of CLOs and the underlying collateral of leveraged loans during past downturns in the economy suggest that CLOs offer an attractive candidature as an asset class for investments on a relative basis. Structural seniority in the capital structure of leveraged loans and the credit enhancements provided by CLOs lead us to lean cautiously in favor of CLOs, as long as CLOs remain CLOs.
3
“Credit Loss Rates on Similarly Rated Loans and Bonds,” Moody’s Investors Service, December 2004.
Please see additional important disclosures at the end of this report.
335
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 50
Lien on Me, Again
June 3, 2005
Primary Analyst: Vishwanath Tirupattur Primary Analyst: Sivan Mahadevan Ajit Kumar, CFA The growth in second lien loan issuance in the last two years has been extraordinary. According to S&P LCD data, second lien loans as a percentage of total institutional loan volume has gone from a low of 0.19% in 2001 to 10.29% by the first quarter of 2005 (Exhibit 1). exhibit 1
Share of second lien loan volume (% of total institutional loan volume)
12% 10.3%
10% 8.0%
8%
6%
4%
3.6%
2%
1.3% 0.7%
0.7%
1.0% 0.3%
0.2%
2000
2001
0% 1997
1998
1999
2002
2003
2004
1Q05
Source: S&P LCD
Second liens as an asset class have also been growing in their relevance to CLO investors. Second lien loans began to appear with notable frequency in CLO portfolios early last year. Given their general classification as senior secured debt, some CLO managers were finding the inclusion of relatively higher yielding second lien loans extremely attractive. However, it did not take long for market participants to become cognizant of the higher degree of risks associated with the higher yields of second lien loans. Rating agencies’ reaction to curb the tendency to capitalize on the broad definition of “senior secured” in CLO indentures has been to require the creation of separate defined buckets for second lien loans. Further, rating agencies have also refined their approach to the assessment of recovery rates to be more reflective of their potential for post-default recovery. Over the last year, we have noticed a steadily rising trend in the size of the second lien buckets in new issue CLOs. From a level of 5% during the middle of last year, it has become increasingly common to have 20% or higher separate second lien buckets in new issue CLO portfolios.
336
Second lien loans are typically subordinated in their rights to receive principal and interest payments from the borrower to the rights of the holders of first lien debt. As a result, second lien debt is clearly riskier than the more senior first lien debt. However, that by itself is no reason to castigate their inclusion within CLO portfolios. Their higher yields do offer a level of compensation for the higher risk. Further, it would be imprudent to clump all second liens into one category. Not only are there significant differences among second lien loans in terms of both risk and reward potential, but also their inclusion has different economic consequences for different tranches of CLOs. In this chapter, we provide a brief background on the distinctive nature of second lien loans as an asset class and their increasing significance to CLOs, and we analyze their impact on the loss distributions of different CLO tranches. ALL SECOND LIEN LOANS ARE NOT CREATED EQUAL
In general, second lien loans have longer maturity and they carry fewer and lessrestrictive financial covenants relative to first lien loans. However, their priority in claims on the underlying collateral vis-à-vis trade and unsecured creditors both before and after bankruptcy, post-petition interest and leverage in reorganization plan negotiations vary widely. In some cases, second lien creditors get a seat at the table alongside first-lien holders, albeit with a lower priority of claims, but in some others they do not. In general, second lien creditors agree by contract to give up some of their rights and defer to the first lien creditors as to the exercise of their rights as secured creditors (“silent” second liens). That said, provisions that define just how silent is “silent” differ. In this context, some of the hotly contested points include the scope and duration of enforcement standstills, waiver of right to vote in bankruptcy, waiver of adequate protection, and waiver of rights to exercise remedies against collateral. The differences in the risk-return profile of second lien loans are evident in the distribution of their spreads. In Exhibit 2, we plot the current spreads of the 58 second lien loans in the S&P LSTA leveraged loan index. exhibit 2
23
Second lien spreads in S&P LSTA leveraged loan index (bp)
11
10 7 5
1 0
900-1000
800-900
700-800
600-700
500-600
400-500
300-400
200-300
100-200
0
1000-1100
1 0
Spread Range (bps)
Source: S&P LCD
Please see additional important disclosures at the end of this report.
337
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 50
It is evident from above that the range of spreads is substantial – from a low of 275 bp to a high of 1,000 bp. Another dataset on spreads – S&P LCD data on a larger universe of second lien loans demonstrates a similar range. As of the end of the first quarter of this year, 55.7% of the outstanding second lien loans were trading at a spread of 500 bp or less and 11.1% were trading at a spread of greater than 850 bp. CLOs are the largest investors of second lien loans. Nearly 50% of all second lien loans issued as of the first quarter of 2005 found their way into CLO portfolios, which are larger than any other class of investors, including hedge funds, insurance companies and prime rate funds. The attractiveness of second liens to CLO managers is evidenced by the spread differential between first and second lien loans. Exhibit 3 shows the average spread of second liens and the gap between the spreads of second and first lien tranches over time. Second lien loans have offered a spread pick up in the 200-400 bp range over the last 18 months. exhibit 3
Average spread of second lien loans and the gap between second and first lien tranches (bp)
800
Second Lien Spread
First Lien - Second Lien Gap
700 600 500 400 300
Apr-05
ME 5/19/05
Mar-05
Jan-05
Feb-05
Dec-04
Oct-04
Nov-04
Sep-04
Jul-04
Aug-04
Jun-04
Apr-04
May-04
Mar-04
Jan-04
Feb-04
200
Source: S&P LCD
MODELING SECOND LIEN LOANS IN CLOS THROUGH RISK-NEUTRAL LENS
Risk-neutral pricing models rely on the notion that the spread at which a particular credit trades reflects the market’s best guess of its level of embedded credit risk. When applied in the context of portfolios of credit risky instruments along with a default correlation assumption, we can generate expected loss distributions for different tranches in a capital structure. These methods are commonly used in the context of synthetic CDOs and other portfolio credit default swaps where the cash flow waterfalls are relatively uncomplicated and the underlying portfolios are static, with no delevering mechanisms and call features. On the other hand, CLO structures contain each of these complications.
338
Still, in the context of evaluating the effect of increasing second lien loan buckets on the expected loss distributions of different CLO tranches, a risk-neutral approach offers useful insights. Specifically, it enables us to evaluate the compensatory effect of the incremental spread pick-up relative to the increase in default risk assumed by the underlying portfolio, as we increase the size of the second lien bucket. It is important to qualify here that we apply risk-neutral models not to “price” the tranches as such, but to analyze the expected losses for different tranches as the proportion of second lien loans in the portfolio increases. In order to clearly distinguish this aspect, we have chosen to tranche the expected losses of a hypothetical CLO portfolio into equal thickness of 10% each such that the attachment and detachment points are spaced as 010%, 10-20%, 20-30% and so on, clearly different from typical CLO capital structures. A HYPOTHETICAL CLO PORTFOLIO
We analyze a hypothetical CLO portfolio with 50 assets consisting of first lien loans and second lien loans. We generated the spreads for these assets drawing randomly from a distribution of spreads with characteristics representative of the spread dispersion of the constituents of the S&P LSTA leveraged loan index as of May 26, 2005. Our assumptions for mean and standard deviation of the different categories of assets in the portfolio are summarized in Exhibit 4. We started with a portfolio consisting of 90% first lien and 10% second lien loans and gradually increased the proportion of second lien loans in 2% increments. Given that the focus of this analysis was on the impact of second lien loans, we did not model other types of assets such as DIP loans, unsecured bonds and the like. In terms of recovery, first lien and second lien loans were modeled at 70% and 40% recovery, respectively. We used a flat correlation of 35%. exhibit 4
Mean Spread Std Dev
Assumptions Used for Generating Hypothetical Portfolio Spreads First Lien BB 246
First Lien B 300
Second Lien 582 138
69
79
Min
138
150
275
Max
513
750
1,000
Recovery assumption
70%
70%
40%
Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
339
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 50
Exhibit 5 shows the expected losses for the first four tranches of the CLO expressed as a % of the size of the tranche, for different sizes of the second lien bucket. As expected, as the proportion of the riskier second loan liens rises, expected losses increase across the board for all tranches. However, the rate of increase is significantly different for different tranches. Following along down the 0-10% column, we see that expected losses for the riskiest tranche increase from 86.9% to 88%, an increase of only 1.25%, as the proportion of second lien loan buckets rises from 10% to 20%. For the 30-40% tranche, they increase from 1.0% to 4.4%, an increase of 341%. For the 10-20% tranche and the 20-30% tranche, the increase in expected losses over the range is 7.25% and 33%, respectively. Impact of Increasing Second Lien Loans
exhibit 5
Tranche Expected Losses/Notional 10-20% 20-30% 60.8% 25.4%
Second Lean 10%
0-10% 86.9%
30-40% 1.0%
12%
87.1%
61.5%
26.9%
1.4%
14%
87.1%
62.1%
28.3%
2.0%
16%
87.3%
62.7%
29.8%
2.6%
18%
87.7%
64.0%
31.7%
3.4%
20%
88.0%
65.2%
33.8%
4.4%
Source: Morgan Stanley
In other words, larger second lien loan buckets do not significantly change the expected losses for the riskiest tranches, but for senior tranches the increase in the second lien buckets can be significant. The rate of increase of expected losses is also significantly different for different tranches. The more senior tranches experience a higher rate of increase of risks as the proportion of the second lien loans in the CLO portfolio increases than the subordinated tranches. The most subordinated tranche experiences the lowest rate of increase in expected losses. Since the benefits of the higher yields of the second lien loans accrue to the equity tranche holders in a CLO in terms of residual distributions (in our analysis, the 0-10% tranche is the closest to a CLO equity tranche), we can surmise that equity tranches are net beneficiaries of the increase of the second lien loan bucket sizes. The increase in the risks of equity tranches is minimal and benefits may be significant. On the other hand, risks of the debt tranches increase as second lien buckets increase in size and they do not receive the benefit of residual distributions. Their pricing should reflect this increased risk exposure. However, whether that has been happening in the CLO market place is unclear. The fact that new issue CLO senior tranche spreads over the last year and half have been tightening is not an adequate demonstration that they have not been. Unlike an economist’s idealized ceteris paribus world, in the real world not all other things remain the same over time. The tightening in CLO debt spreads, despite their increased risk exposure due to increasing second lien loan buckets is not as much an aberration as it might seem and is broadly consistent with the all around declines in risk premiums.
340
There are several important caveats of this analysis that deserve reiteration. First, most CLOs will have significant reinvestment of collateral during their life, thanks to the short maturity of leveraged loans and the potential for prepayments. Since the uncertainty about the level of spreads at which future reinvestments can be made, the potential for reinvestments can only be construed as a risk not modeled in our analysis. Second, as noted earlier CLO structures have complex waterfall mechanisms and triggers that redirect cash flows to senior tranches at the cost of subordinated tranches as the portfolio experiences stress. Such triggers are inherently difficult to model in this framework. Third, CLO portfolios are not static portfolios and managers make discretionary changes in the portfolio, which can have meaningful impact. Finally, as we point out in Chapter 70, for risk-neutral analyses of the sort we have used here, to replicate market prices of CLO debt tranches, particularly senior parts of the capital structure, it takes extremely high correlation and extremely low recovery assumptions. In other words, using “reasonable” correlation assumptions CLO debt tranches appear to be priced in the market significantly wider than implied by our risk-neutral models. It can be argued that the debt investors are anyway being compensated handsomely, notwithstanding the incremental risk of second lien loans. CONCLUSION
We revert to the point we made at the outset. Second lien loans are here to stay and will remain significant to CLO portfolios. There are significant differences in risks and rewards resulting from their inclusion and increase of their share in CLO portfolios. They affect different tranches very differently. The most subordinated tranches reap the benefits of their higher spreads but do not suffer a significant increase in the negative consequences of their inclusion. The more senior tranches are likely to experience higher expected losses. Whether the debt tranches are adequately being compensated for the incremental increase in risk is yet unclear.
Please see additional important disclosures at the end of this report.
341
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 51
DAL + NWAC = Correlation Without a Model
September 23, 2005
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj If you have ever wondered what default correlation really means, we have a very simple non-mathematical answer: Delta Air Lines and Northwest Airlines filing for bankruptcy within minutes of one another is a real-world example of default correlation. We have often stated that the correlation values implied by today’s tranche pricing models are a difficult concept to measure because actual default correlation is not easily observable in the marketplace. The argument goes that this difficulty in observing realized correlation (unlike realized volatility, which is indeed observable) makes it hard to gauge ‘fair value’ for instruments whose value is affected by correlation of the underlying assets. Well, at least in some respects, the September 14th bankruptcy filings by two major US airlines provide the market with very much a realworld example of how to think about correlation risks. Yet, more importantly, the near simultaneous bankruptcy filings reveal the dynamics that drive the actions of companies near the bankruptcy threshold, which may be less about industry sectors and more about other common problems, like unfunded pension liabilities, labor costs, healthcare issues and negative exposure to rising commodity prices. As such, it is far from pure chance that these two companies chose to file for bankruptcy protection so soon after one another. However, going from these realworld events to correlation model parameters is a less than clear process. On a forward looking basis, there are two important insights to gather from the bankruptcy activity in the airline sector. First, it is not necessarily a sector issue that drives default correlation higher, as many of the problems in the airline sector also appear in the auto and auto parts sectors, and have historically appeared in other industrial corners of the credit markets. Second, higher default correlation among a few credits means that these credits are actually less correlated with the rest of the market (or a large portfolio), which can explain why we have seen implied correlation actually fall in the structured credit markets recently. This may feel counterintuitive, but we attempt to clarify it with a simple example. LEGACY CARRIERS ARE THEIR OWN ANIMAL
While the combined impact of such macroeconomic events as rising oil prices, terrorist events, and an investment environment unfriendly to holders of pension liabilities created a perfect storm for the US legacy carriers, it is worth noting that many international and low-cost providers in the industry are still able to survive, if not thrive, under these conditions. The problems facing the legacy carriers are very specific to the business structure of these carriers. That structure may be more similar to the structures of companies outside the airline industry than we might expect.
342
exhibit 1
Moody's 5-Year Cumulative Default Rate – IG Universe
2.5%
2.0%
1.5%
1.0%
0.5%
2000
1998
1996
1994
1992
1990
1988
1986
1984
1982
1980
1978
1976
1974
1972
1970
0.0%
Cohort
Source: Morgan Stanley
UNDERLYING CAUSES OF CORRELATION
In some respects, the cyclicality/correlated nature of default rates can be examined from the perspective of exogenous events that cause defaults or bankruptcy filing in particular circumstances. Consider the most recent peak in the default cycle as illustrated in Exhibit 1. The last spike (2000-2001) in default rates was driven by some very sector-specific events like the telecom overbuilding and subsequent bust. Clearly, this was an industry-specific event. Other seemingly unrelated bankruptcies were actually driven by the broader theme of fraud, which some have argued was based on the pressure to manage/manufacture earnings on the back of investor growth expectations during the late 1990s equity bubble. The extreme focus on the potential for fraud immediately after the bubble burst made fraud more likely to be discovered and hence led to defaults across several sectors. Recent events in the airline space also seem to be industry specific, initially, but they could also point to a broader dynamic across many industries. While the airline industry has obviously struggled, equally important for both Delta and Northwest is the status of their pension plans and their ability to address these and other labor cost issues going forward. The pending deadline for bankruptcy law changes may have served as the catalyst to bring these issues to the fore. When examined through this lens, the pension and other labor cost issues affecting the airline industry might expand and could prove to be a driver in defining the level to which defaults across other sectors are, in fact, correlated. In Exhibit 2, we show an updated list of companies with the highest unfunded pension obligations relative to their equity market value. The original was published by our colleague and high yield strategist Brian Arsenault (see High Yield Strategy Insights, 2/14/05, “Budgeting for Pension Reform”) and contained Collins & Aikman Corp., as well as Delta and Northwest Airlines. We would argue that this list highlights a set of companies that are likely to have high realized default correlation and also have the potential to be in a financial position where this common theme could be the deciding factor as to whether these companies default or not. A quote from Robert “Steve” Miller, CEO of Delphi Corp., in a recent Bloomberg article illustrates the point: ``What happened in steel four years ago, what happened in airlines over the last couple of years, is coming to this town.''
Please see additional important disclosures at the end of this report.
343
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 51
exhibit 2
Issuer 1 Delta Air Lines Inc 1 Northwest Airlines Corp Exide Technologies Delphi Corp Continental Airlines Inc AK Steel Holding Corp AMR Corp Goodyear Tire & Rubber Co Hayes Lemmerz Int’l Inc Dura Automotive Systems Inc Visteon Corp Ford Motor Co Austral Pacific Energy Ltd General Motors Corp ArvinMeritor Inc Unisys Corp Navistar International Corp Smurfit-Stone Container Corp Abitibi-Consolidated Inc Tenneco Automotive Inc Bowater Inc PolyOne Corp Dana Corp Hercules Inc TRW Automotive Holdings Corp Terra Industries Inc Timken Co Alcan Inc
Not Only One Sector – Unfunded Pension Liabilities Relative to Equity Market Capitalization Ticker DAL NWAC XIDE DPH CAL AKS AMR GT HAYZ DRRA VC F AEN GM ARM UIS NAV SSCC ABY TEN BOW POL DCN HPC TRW TRA TKR AL
Industry Group Airlines Airlines Auto Parts & Equipment Auto Parts & Equipment Airlines Iron/Steel Airlines Auto Parts & Equipment Auto Parts & Equipment Auto Parts & Equipment Auto Parts & Equipment Auto Manufacturers Oil & Gas Auto Manufacturers Auto Parts & Equipment Computers Auto Manufacturers Forest Products & Paper Forest Products & Paper Auto Parts & Equipment Forest Products & Paper Chemicals Auto Parts & Equipment Chemicals Auto Parts & Equipment Chemicals Metal Fabricate/Hardware Mining
Funding Gap/ Mkt Cap 1235% 875% 266% 256% 215% 147% 145% 112% 102% 81% 73% 68% 62% 43% 40% 36% 34% 33% 33% 32% 29% 27% 25% 22% 22% 22% 29% 24%
1
Market Cap based on the average of the 6 months prior to the bankruptcy filing. Source: Morgan Stanley
One point to make very clear is that the extent to which pension and healthcare issues are addressed through legislation does not reduce the correlation these companies are likely to experience. In fact, their exposure to that exogenous risk is a further element of the hidden links among them. If pension reform legislation passes, they will all be better off and to the extent the issue is not addressed, they will all be worse off. Either way, their fates are tied together by the invisible threads of the underlying problems facing our private healthcare and retirement system. VISUALIZING CORRELATION
It should immediately jump out to those familiar with the standard structured credit relative value tools that recent implied correlation moves seem counterintuitive given the dynamics we just described. At a time when we see some very highly correlated default events (and the potential for other events driven by the same fundamental factors) the implied correlation of the most junior tranches has declined.
344
These two dynamics are not actually inconsistent but this is where the critical assumptions made by market standard models show their spots. The seeming contradiction is best illustrated in an example, which we show in Exhibit 3. Consider a world in which the likelihood of any one of one hundred credits defaulting is equally related to any other or equivalently they all have the same pair-wise correlation. Now consider a world in which the default propensities of a small subset (three, in this example) of these companies are very highly correlated to each other but not to the rest. These are equivalent to World A and World B in the exhibit, respectively. The average correlation at the far right of the exhibit is something like the implied correlation we observe in structured credit markets today – it summarizes the aggregate level of correlation but it misses the fact that many of the names have very different relationships than this “average” indicator. The high correlation of a few names can be as important a driver for the most subordinate tranches as the correlation among the rest of the names. Thus, while we may observe a lower implied correlation, the market could actually be pricing a scenario of high correlation for a small subset of companies. Correlation – Details Matter
exhibit 3
Credit 1 2 3 4 5 … 100
Credit 1 2 3 4 … 100
1 50% 50% 50% 50% … 50%
1 90% 90% 10% … 10%
2 50% 50% 50% 50% … 50%
2 90% 90% 10% … 10%
3 50% 50% 50% 50% … 50%
3 90% 90% 10% … 10%
WORLD A 4 5 50% 50% 50% 50% 50% 50% 50% 50% … … 50% 50% WORLD B 4 5 10% 10% 10% 10% 10% 10% 50% 50% … 50% 50%
… … … … … …
100 50% 50% 50% 50% 50%
Average 50%
… … … … …
100 10% 10% 10% 50%
Average 48%
Source: Morgan Stanley
CORRELATION: THE REAL-WORLD KIND
The recent filing of NWAC and DAL might be the proverbial shot fired over the bow for investors in companies with outstanding pension funding and labor cost issues. For those with exposure to these names in structured credit products, the common themes affecting the fates of these companies have important relative value implications and could be the source of both risk and opportunity. It seems prudent to keep an eye on correlation, the real-world kind, that is.
Please see additional important disclosures at the end of this report.
345
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 52
What Does “Bespoke” Really Mean?
September 30, 2005
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Investors tend to think of today’s structured credit market along two dimensions: the well-defined, liquid and transparent index tranche market, and the much more fragmented market for one-off or bespoke structures. Index tranches and the related hedge fund activity are an important flow today and have been responsible for moving the structured credit business out of its niche into the mainstream. However, much of the risk held by end-investors is in bespoke form. In fact, the history of structured credit is very much about investing in bespoke opportunities, and we argue that its future is, too.
© 2005 Morgan Stanley
346
The process of investment grade credit investing may have reached an important crossroads today, for two general reasons. First, policies and guidelines from Basel II to the recent FASB proposal are generally very supportive of using structured credit vehicles in the investment management process. Given the importance of credit ratings in both contexts, the global pool of credit investors may find that cash and synthetic structured credit provide the best yield for the rating and can be very efficient from regulatory capital and income volatility perspectives. Second, consolidation among money managers has created extremely large corporate bond portfolios, and we have not yet met a portfolio manager who feels that navigating such a portfolio through the credit cycle is an easily feasible task. Bespoke structured credit solutions, in long or short form, may help the cause. So what does “bespoke” really mean? In credit market circles, we argue that the term’s prevalence in American English is related to the US credit environment becoming more inviting to structured credit solutions, in many ways taking a page from Europe. In the remainder of this chapter, we focus on the investing environment and risks for using structured credit for three specific communities of investors, namely banks, insurance and asset managers. BANKS AND BASEL II
Although technically not implemented until the end of 2006, Basel II has already had profound implications on how global banks are approaching investing in the credit markets. We defer all detailed questions on Basel II to our esteemed global bank analyst team, led by Jackie Ineke, but there are a few points we feel comfortable making here. 1 Today (i.e., under Basel I), corporate bond investments are 100% risk-weighted, as are tranches of cash and synthetic CDOs. Under Basel II, these risk weightings will fall dramatically in general, providing much regulatory capital relief to banks (see Exhibit 1). Synthetic CDOs that are funded and rated are treated like other securitizations (e.g., ABS) and have similar risk weightings. Although not a major issue, we note that risk weightings for below investment grade rated tranches are actually higher under Basel II. exhibit 1
Mean Spread Std Dev
Basel II Risk Weightings Favor Structured Credit First Lien BB 246
First Lien B 300
Second Lien 582 138
69
79
Min
138
150
275
Max
513
750
1,000
Recovery assumption
70%
70%
40%
Note: Assumes RBA approach. Source: Morgan Stanley, Basel II
1
For more information on Basel II issues related to credit derivatives and CDOs, see “Basel II – Credit Derivatives and Risk Mitigation,” October 1, 2004, and “Basel II – ABS,” September 13, 2004.
Please see additional important disclosures at the end of this report.
347
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 52
The market impact of Basel II is important, in our view. Banks have been fairly significant participants in the cash and synthetic structured credit markets over the past two years, mainly in the senior parts of the capital structures. We would argue that this activity is very much linked to Basel II. INSURANCE COMPANIES AND FASB
The following information contains a general, summary discussion of certain select accounting issues. Any such discussion is necessarily generic and may not be applicable to or complete for any particular investor’s specific facts and circumstances. Morgan Stanley is not offering and does not purport to offer accounting advice and this information should not and cannot be relied upon as such. Morgan Stanley has prepared this information based on our understanding of the issues following review of materials prepared by third party accounting experts. The positions of such third-party experts may be reasoned and the views of other third party experts may differ from those summarized herein. Potential investors are urged to consult their own accounting advisors before making any investment decisions regarding any transaction. While it is difficult to compare Basel II and its complexity to anything else in the world, the ramifications of Basel II on structured credit markets have some similarities to what could happen in the US with insurance companies and the recent FASB proposals. The FASB discussions in the marketplace are related to a proposed amendment to FAS 133 and 140. 2 The statement proposes to account for credit-linked notes (with embedded derivatives) issued from a special purpose entity (SPE) similar to conventional cash corporate bonds. This could be interpreted as implying that the embedded credit derivative in such a structure (e.g., a single-name CDS or a tranche of a synthetic CDO) would not be bifurcated (as is common today) and therefore would not be subject to mark-to-market treatment if the investor avoided consolidating the SPE issuing the credit-linked note. To avoid consolidation, the SPE would have to be static or rules based (QSPE, like many are), or actively managed but requiring the investor to own a maximum of 50%. For the credit investors in insurance companies (and other institutions where bonds and derivatives have different accounting treatment), the recent FASB-proposed language may represent a secular shift in the investment management process. The taking of credit risk in synthetic form to date has largely been constrained in this community because mark-to-market treatment would introduce income statement (and thus earnings) volatility. In fact, during the 2002-2003 period, we saw many insurance companies exit the structured credit space for precisely this reason. The new FASB language has the potential to make both synthetic and cash instruments equally as attractive to this community of established corporate bond investors. The credit market implications of this potential secular shift should not be ignored. It can do everything from compress the basis in single names to force a rally in mezzanine and senior tranches or motivate further financial engineering. If there is a preference for taking credit risk in structured credit form, what parts of the markets could suffer? In the near-term, we do not think anything will, as there is still plenty of 2
See “FASB Exposure Draft of Proposed Statement of Financial Accounting Standards, No. 1210-001 (August 11, 2005).”
348
investing capacity in the market; however, over time, assets with the least yield for the rating would become less attractive, particularly investment grade corporate bonds. ASSET MANAGERS AND SIZE
In the traditional unlevered asset management world (mutual funds and separate account management), there has been a slow but steady acceptance of using credit default swaps. From our experiences, one of the biggest issues this community has is size. Consolidation and a general growth of assets under management have resulted in very large corporate bond portfolios. Using single-name instruments (including the popular CDX indices), it is extremely difficult to manage these portfolios in any macro form, as the single-name markets lack the liquidity and the technology to trade large quantities of credit on the proverbial wire. Bespoke structured credit solutions, from a long or short perspective, may be helpful to solve these problems, as such structures can allow for efficient transfers of large amounts of credit risk, and even active management of the specific risk after the trade is done. There are also idiosyncratic versus systematic credit risk approaches investors can take, which can be quite useful in the macro context. Our discussion of these types of solutions with investors has been limited, but as this community has already spent a good amount of time attempting to cross the derivatives hurdle, structured credit solutions could be easily applicable. REAL WORLD VALUE IN STRUCTURED CREDIT
Real world default experience (which is generally lower than market-implied) provides a strong argument for taking mezzanine type risk through structured credit instruments on investment grade portfolios. Yet, a key issue is sampling risk (see Exhibit 2). While expected losses based on historical data appear to rarely penetrate mezzanine tranches, the fatter tails resulting from sampling risk in smaller portfolios show the risk to which most mezzanine investors are exposed. In either case, the expected 5-year loss for this portfolio and universe is 1.0%. exhibit 2
Sampling Risk: Smaller Portfolios Have Bigger Tails
45% Portfolio (125 Names) 40%
Universe (1,000 Names)
35% 30% 25% 20% 15% 10% 5% 0% 0.000.25%
0.250.75%
0.751.25%
1.251.75%
1.752.25%
2.252.75%
2.753.25%
3.25- 3.75% 3.75% or More
Note: 5-Year period, investment grade universe, CDX IV ratings distribution. Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
349
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 52
Rolling the clock back five years, we consider the impact of tail events on performance. Consider a hypothetical transaction with 250 names and unlucky enough to have exposure to 14 fallen angel defaults. We estimate that such a transaction would have realized losses of approximately 2.8%, not even penetrating a 3% attachment. If we then consider a transaction with only 100 names, but still containing all the same defaulted names, losses would be approximated to be about 7%, almost completely running through a 3-7% tranche. Uncertainty around recovery also has some important implications. The above 100-name transaction would have realized losses of 11%, if realized recoveries had averaged 20%, easily penetrating both a 3-7% and a 10% attachment. While average recovery assumptions are likely reasonable, the risk in many structures comes from the fact that a particular portfolio could have recoveries that vary significantly, another dimension of the sampling risk associated with structured credit products. We note that if the same transaction experienced average recoveries of 80%, we would approximate losses to be the very manageable 1.8%, not penetrating any tranche with a modicum of subordination. Critically important to the performance of the tranches are both forms of sampling risk: realized default and realized recovery rates, which can vary dramatically and have a direct impact on the erosion of subordination. The market appetite for this combination of risks, both likelihood of default and recovery volatility, is what the tranched credit events of May 2005 are all about. Recent market dynamics illustrate the point. While equity and super senior tranches have traded with a fair amount of volatility, the mezzanine and senior parts of the capital structure have been much less volatile, highlighting the strength of the hands in the investment community that buys this risk. Although risk premiums have essentially been transferred between the riskiest and the safest parts of the capital structure, the middle has been remarkably stable. Why is this important? If one wonders what could slow the huge potential structured credit flow we have described above, a few (or even a single) extreme events during a negative credit cycle could be responsible. We saw many people questioning their structured credit investments during the last cycle after two unexpected very low recovery investment grade defaults. While this is natural, index tranche trading in recent months tells us that the market has matured a bit and seems to better differentiate between very idiosyncratic events and those affecting the broader market. Whether the market does so in the future remains to be seen.
350
MARKET IMPLICATIONS – WHERE’S THE BALANCE?
Basel II, FASB proposals and the reality of managing large corporate portfolios has meaningful implications on the credit markets. As both Basel II and FASB are supportive of taking credit risk in structured form, leaning against this force over long periods of time is a tough trade. However, there will be some balance in the markets, as Basel II also makes carrying hedged positions more efficient for banks, and the structured credit flows from the asset management community could easily be long or short. In any event, these changes in investment style are tantamount to a secular shift in the portfolio management process. In some sense, the structured credit activity in May, when equity tranches massively underperformed mezzanine tranches, demonstrates the powerful force of investors in the insurance and banking communities versus hedge funds and dealers. Could a negative basis be the norm based on increased structured credit activity? More likely the basis will flip-flop based on the direction of the largest flows. In the future, most investors will not be limited solely to risk-taking activity as it relates to structured credit; they will also shed risk using these tools. This, in turn, could lead to a much healthier structured credit market across the capital structure.
Please see additional important disclosures at the end of this report.
351
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 53
Delphi – Let’s Step Outside and Settle This
October 14, 2005
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Vishwanath Tirupattur Primary Analyst: Viktor Hjort Primary Analyst: Afsheen Naqvi The well-telegraphed Delphi bankruptcy filing is an important event for credit derivative and structured credit markets, impacting both operational activity and market pricing in nearly every corner of the market (see Chapter 19 for our warning flags). The bankrupt auto supplier represents the first meaningful US fallen angel default (within one year of an IG rating) since 2002, when about a dozen such defaults caused havoc in the thenburgeoning structured credit markets. Yet, the market is very different today than in 2002, for many reasons. Market pricing in mezzanine tranches tells us that Delphi is being considered an idiosyncratic event, even though its default is “correlated” with the Delta Air Lines and Northwest Airlines bankruptcies, given similarity of issues and the approaching bankruptcy reform deadline (see Chapter 51). In 2002, we could not have made the same statement, as the market pricing compass then unambiguously pointed in the systemic direction. However, Parmalat, a low-recovery late 2003 European fallen angel default, had large exposure in a structured credit world that more closely resembles today’s (including index tranches, see Chapter 60). There is clearly a lot to focus on with respect to the Delphi bankruptcy. Dealers each have thousands of CDS trades to settle (compared to perhaps a hundred or so during the larger 2002 bankruptcies), yet tri-party systems will ease the netting process. While it is natural to ask how much derivatives risk is outstanding in Delphi, the answer to that question is actually only partially relevant. Equally important is understanding the settlement process, and any potential imbalances that could occur due to unconfirmed trades, delivery failures and risk management issues. Market participants are working on a protocol, similar to the Collins & Aikman protocol earlier this year, although the details are not yet clear. The biggest unknown at this point is the range of Delphi recoveries over this settlement period, which is very much related to potential imbalances described above, as well as to the protocol. This recovery uncertainty is a key risk, influencing both P/L and ratings actions. Delphi bonds are trading in the high 50s now, which makes it quite different from many of the low recovery defaults of 2002 and Parmalat.
352
exhibit 1 Year 2002
2003
2004 2005
Fallen Angel Defaults (Within One Year of IG Rating)
Company AT&T Canada Banco Commercial S.A. Covanta Energy Corporation Duty Free International, Inc. Energy Group Overseas Intermedia Communications Corp. Kmart Corp. Marconi Corp., Plc MCI Communications Corp. NRG Energy, Inc. Petroleum Geo-Services ASA PG&E National Energy Group Qwest Capital Funding, Inc. Teleglobe Inc. TXU Eastern Funding Company WorldCom, Inc. British Energy Plc Northwestern Corp. Parmalat None Delphi Corp.
Source: Morgan Stanley, Moody’s
TRANCHE RATINGS IMPACT IS KEY TO WATCH
One of our biggest focuses is determining the impact of the Delphi bankruptcy on tranche ratings, as downgrades can drive flows from end investors. In its initial comments, S&P stated that Delphi appears in over 800 synthetic CDO transactions, representing about 35% of its universe. S&P estimates that at least one tranche in one-third of these transactions will be downgraded. Moody’s has announced that it has identified over 300 CDO transactions with Delphi exposure, including 71 cash CDOs, and that it expects some ratings impact even though it is too soon to determine the extent. Fitch has placed 73 tranches from 40 CDOs on watch negative, of the 136 transactions within their rated CDO universe with exposure to Delphi. Transactions with Delphi exposure but not placed on watch negative are expected to have sufficient credit enhancement to absorb the loss, assuming a 5% recovery rate, which is rather extreme. Several factors would likely determine rating agency actions: size of the exposure to Delphi, previous subordination erosion, deterioration of credit quality elsewhere in the portfolio, capital structure, leverage achieved within the structure and the final recovery amount. We argue that final recovery values will be the most consequential drivers of the ratings actions. Using S&P's publicly available CDOEvaluator model, we find that for a simple portfolio (we use DJ CDX 3) and structure, the ratings of the mezzanine tranches (BBB and below) are most at risk for downgrade actions. The recovery rate assumption used matters, and higher recoveries will likely result in fewer tranches being at risk for downgrade.
Please see additional important disclosures at the end of this report.
353
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 53
Prior to default, Delphi was a CCC- credit, which, itself, has a very high probability of default, so an actual default does not necessarily introduce a lot more risk into the structure; furthermore, a high recovery default (relative to recovery assumptions) can actually reduce risk in the structure, ironically. Yet, agencies do not always operate in a continuous fashion, so the Delphi bankruptcy event may force analysis of all deals with Delphi exposure. COMPLEX STRUCTURES, DEAL RESTRUCTURING
Delphi appears as a popular credit in many of the market’s more complicated structures, including CDO-squareds where fixed recovery trades are also common. These fixed recoveries are typically much lower than current market prices for Delphi bonds (3540% fixed recovery versus high 50s for the bonds), which puts these transactions at greater risk of a downgrade. Delphi overlap in these CDO-squared structures is also an important risk factor, but this is a double-edged sword, serving both to reduce and raise exposure (depending on the deal), relative to a simple CDO structure. Ratings downgrades do not necessarily bring in forced sellers of credit risk, even among the most ratings sensitive investors. It is not uncommon for CDO tranches to be restructured to maintain or raise ratings through a variety of mechanisms, a common event three years ago. DELPHI’S IMPACT ON THE STRUCTURED CREDIT MARKET?
We have stated numerous times in the past that it will take actual defaults to derail the structured credit bid, not just headlines or spread widening. Is Delphi enough to force this? Based on feedback we have received from numerous investors globally, the early read is “no,” although we caution that we are gathering this feedback ahead of the Delphi settlement in tranches and potential ratings actions. We believe that near-term flows will slow as investors grapple with operational issues (i.e., every investor will receive credit event notifications, regardless of subordination) and explain potential ratings actions internally. From a ratings perspective, we would argue that mezzanine notes falling below investment grade are one potential trigger point for flows, as are significant markto-market losses. It is also important to note that many European transactions are managed in nature, and it is indeed possible that managers traded out of Delphi this year, although most likely at a loss to where they entered, so ratings and mark-to-market impacts may have already been felt. Yet, the secular shifts that have made structured credit an important solution for many European investors have by no means gone away, so we expect flows to be healthy in the medium term (for our sermon on this topic, see Chapter 52). In particular, Basel II makes high-rated synthetic CDOs one of the key beneficiaries of regulatory capital relief. Given the importance of the bank sector among European structured credit investors, this has already had a major impact on tranche pricing, even if not implemented for another year (please refer to Morgan Stanley European banks analyst Jackie Ineke’s report “Basel II A-Z,” November 15, 2004). However, ratings are not the only determinant of flows, and the European investor base is gradually becoming driven more by mark-to-market than by ratings, at least in the banking sector, given IAS 39 accounting standards. This, however, is not inconsistent with the current technical environment. Spread volatility at the senior level has
354
remained very low, and Delphi has not changed this. The key uncertainty at this stage relates to any follow-on effects on – for instance – GM. The crucial difference is on the forward-looking view: Parmalat defaulted in a high but stable correlation environment. The Delphi filing takes place in an environment where tranche markets allocate a much higher (and rising) proportion of risks to equity tranches, suggesting the market expects risks of a limited number of losses to pick up. HYBRID SETTLEMENT – THINKING ABOUT THE ECONOMICS
What is different this time versus the prior instances of fallen angel defaults in the US is that we now have a hybrid settlement mechanism in place for the standard index tranches. While the entire process may seem convoluted at first glance, once we consider the needs and motivations of the various counterparties in the tranched markets, it begins to seem more reasonable. Essentially, the hybrid process involves the following steps for all tranches regardless of whether the loss penetrates the tranche. The buyer of protection identifies the amount of bonds (up to the maximum per name exposure of the position) they intend to physically deliver to the seller, who conducts an auction for the bonds. The price of the auction is used to define the amount of subordination lost for the tranche, as well as any loss payments, if the tranche has been penetrated. The seller pays the market price for the delivered bonds (and presumably can deliver them into the auction for the same price) and, if the tranche is penetrated, pays par minus this market price (the recovery rate) for the full per name notional exposure. WHY CREATE ALL THESE HOOPS TO JUMP THROUGH?
Consider the simple and fairly typical example we show in Exhibit 2. In this example, the buyer of mezzanine protection (Counterparty 1) has done so versus a “delta” position (6x) in the index, while the seller of protection (Counterparty 2) has done so outright. Further, consider the events following a portfolio name default. Counterparty 1 will receive approximately $1.9 million par of bonds from their index “delta” hedge, which is generally physically settled, at least officially. Because of the hybrid settlement process, these bonds can then be physically delivered into the tranche hybrid settlement to set the amount by which the subordination is reduced on the tranche position. While the buyer is free to vary the amount of bonds actually used in the process, the hybrid settlement provides a means to directly link the disposal of bonds delivered through a single name or index “delta” hedged position and subordination loss in a tranche position. All in all, the process is designed to prevent gaming of the auction by either counterparty: the buyer defines the amount of bonds physically delivered into the contract while the seller has the ability to affect the outcome of the auction by participating. The former directly affects the cost of any potential gaming on the part of the seller created by the later, with the goal being to reach a fair price for the defaulted bonds.
Please see additional important disclosures at the end of this report.
355
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 53
Delta Hedged or Not? Two Sides of Settlement
exhibit 2
Counterparty 1
Counterparty 2
Buy/Sell Protection Buy
Tranche CDX 3-7%
Notional 40,000,000
Per Name Notional Exposure 8,000,000
Sell
CDX Index
240,000,000
1,920,000
Sell
CDX 3-7%
40,000,000
8,000,000
Source: Morgan Stanley
GLOBAL VS. INDIVIDUAL SETTLEMENTS – HANGOVER ON THE HORIZON?
While the Collins & Aikman settlement provides us with an example of how the investor and dealer communities can cooperate to keep markets healthy and orderly, Delphi’s default will be a test of the incentives of all the parties involved because of the sheer amount of risk outstanding. The motivation for a global settlement process for the index products is clear. Contracts that go through a global process will remain fungible. This is true because the loss in subordination for all these tranches will be universally consistent, like the HY CDX 3 and 4 tranches today. All of these tranches had their subordination reduced by 0.56375% based on the Collins & Aikman auction on June 14. If each contract were settled via a different auction with different prices defined for the purposes of both loss payments and subordination erosion, the market would be faced with tranches that not only differ from other tranches that were once identical but also tranches that are not necessarily contiguous with tranches that are junior or senior in the capital structure. This will create situations where two tranches that were once fungible (two 0-3% tranches for example), become marginally different, say, one 02.55% tranche and one 0-2.52% tranche. A key element to this risk is the fact that the last investment grade index to contain Delphi was actually DJ CDX 3, which is already two generations off the run, so future liquidity may not be as important a consideration as if the current on-the-run index contained the defaulting name, as was the case in the high yield tranches and Collins & Aikman (as well as Delphi). As discussed in the previous section, the potential for volatility in recovery rates during the settlement period remains significant, particularly in bespoke tranches that may not be part of any global protocol. To the extent bespoke investors are hedged with standard index tranches, a hedge mismatch may ensue because of the difference in the timing and nature of the settlement. Also, the settlement process for bespoke transactions may be different from standardized single-name CDS, indices, and index tranches. LOOKING FOR RELATIVE VALUE
The current pricing on the 3-7% tranches of the DJ CDX series 3, 4 and 5 provides interesting insight into how the market is pricing risk (see Exhibit 3). While the CDX 3 37% tranche will suffer some loss in subordination due to Delphi’s default (0.4% based on a 50% recovery), Delphi is not in the CDX 4 or 5 portfolios – yet the tranches price almost flat (CDX 4 actually trades wide) to one another, mid-market. The implication is that the market values the incremental risk associated with the CDX 4 or 5 positions at about the same value as the impending loss in subordination for the CDX 3 position.
356
exhibit 3 Series CDX 3 (w/Delphi) CDX 4 CDX 5
Maturity 3/20/2010 6/20/2010 12/20/2010
3-7% Tranche Pricing – Which One Seems Wrong? Bid Spread 112 130 122
Offer Spread 132 134 127
Mid Spread 122 132 125
Source: Morgan Stanley
The incremental risk can be divided into two components: the differing portfolio composition (see Exhibit 4), and the differing maturities (see Exhibit 3). Essentially, the market is saying that for the 3-7% tranches, the risk associated with the three differing names in CDX 4 (versus 3) and the 3 month of longer maturity are worth about 10 bp more than the impending loss in subordination in CDX 3. The market is also saying that the extra nine months of maturity and the 11 different names in the CDX 5 portfolio are worth about the same as the loss of subordination in CDX 3. Given the portfolio differences between CDX 3, 4 and 5, buying protection on the 3-7% tranche of CDX 3 and selling protection on the 3-7% tranche of either CDX 4 or 5 is appealing to us. exhibit 4
CDX Composition Differences
In CDX 3 and Not in CDX 5 Bombardier Inc. Delphi Corporation Eastman Kodak Company Ford Motor Credit Company General Motors Acceptance Corp. Kerr-McGee Corporation Liberty Media Corporation May Department Stores Company Maytag Corporation MBNA Corporation Sears Roebuck Acceptance Corp.
In CDX 5 and Not in CDX 3 American Axle & Manufacturing, Inc. The Gap Inc. Hilton Hotels Corporation IAC InterActive Corp. Knight Ridder Inc. Limited Brands Inc. Marsh & McLennan Cos. Inc. Radio Shack Corporation Sabre Holdings Corp. Sara Lee Corp Toll Bros. Inc.
In CDX 3 and Not in CDX 4 Bombardier Inc. Delphi Corporation May Department Stores Company
In CDX 4 and Not in CDX 3 American Axle & Manufacturing, Inc. Hilton Hotels Corporation Lear Corporation
Source: Morgan Stanley
USEFUL MARKET EXPERIENCE – WHAT TO WATCH
The Delphi bankruptcy is an important experience gathering opportunity for a structured credit market that lacks a lot of grey hair. In our view, a high recovery value over the next month will limit ratings downgrades relative to the early statements by the agencies, and investors still have a strong motivation to establish a global settlement protocol, even though it does not affect the market’s most liquid on-the-run products. Three things to watch over the next few weeks: what the rating agencies actually do, where Delphi bonds trade as an indicator of potential imbalances, and what happens to GM. All three have meaningful implications on subsequent credit market flows.
Please see additional important disclosures at the end of this report.
357
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 54
SF CDOs: ABSolutely Subject to Change
October 21, 2005
Primary Analyst: Vishwanath Tirupattur Primary Analyst: Sivan Mahadevan The emergence of structured finance (SF) CDOs as the largest segment of the cash CDO market is a seminal development in the annals of the structured credit market. From nearly insignificant levels in 2000, today’s SF CDOs, composed of collateral from RMBS, CMBS, REIT debt, ABS and CDO sectors, account for over 40% of all new issuance of cash CDOs, a growth rate unmatched by any other CDO sub-sector. Within the reasonably short time that they have been in existence, SF CDOs have undergone significant changes in transaction structures, underlying collateral composition and currency denomination. An appreciation of the factors that have contributed to this growth and understanding them from structural and collateral composition perspectives is important to current and potential market participants. In this chapter, we provide an overview of the structured finance CDO market, review important developments in the market and discuss the risks associated with the product A CHANGING LANDSCAPE
The growth in SF CDOs has been impressive, more notably during the last two years (Exhibit 1). Based on our data, the 2004 issuance was nearly twice the 2003 issuance level and within the first nine months of this year, issuance is already ahead of 2004 full year levels by about 28%. exhibit 1
Growth in Structured Finance CDO Issuance
$Billion 35
Mezzanine ABS
High Grade ABS
Real Estate
Other
31
30 25
25 20
20 15
14
13
12
11
11
10 5
24
7 5
4 0.1
1
2
3
3 4
5
0.1
0 2001
2002
2003
2004
2005
Source: Morgan Stanley, Moody’s, S&P, Fitch IBCA, MCM Corporate Watch, Asset-Backed Alert, Intex
We categorize the SF CDO universe into four broad categories based on the underlying collateral and credit quality. We classified transactions with average underlying portfolio credit quality of A3/A- or below at deal inception as mezzanine ABS and those with higher average credit quality as high grade. While the underlying portfolios of both mezzanine and high grade transactions frequently include CMBS/REIT debt, we categorized transactions with collateral consisting entirely of CMBS/REIT debt
358
separately as real estate CDOs. 1 Finally, we grouped transactions whose underlying collateral consists predominantly of CDO securities or synthetic securities under the generic “other” category. All categories of SF CDOs have grown significantly. Mezzanine and high-grade transactions dominate, accounting for over two-thirds of the SF CDO universe. Real estate CDOs issuance has nearly doubled between 2004 and 2005. Due to the private nature of synthetic transactions, including single tranche synthetics, accurately tracking synthetic SF CDO issuance proves difficult. While our new issuance-tracking database has improved over time, we think we are somewhat underestimating the synthetic ABS CDO issuance volume. The total volume of 2005 issuance for the first nine months of the year totaled $79.7 billion from over 160 transactions. Average deal size has increased in both high grade and mezzanine transactions mainly due to ABCP tranches and other structural efficiencies in the senior parts of the capital structure. 2 Shifts in the currency composition of SF CDO issuance are noteworthy. In 2001, all of the visible issuance was USD denominated. While in 2005, USD denominated issuance still dominates with about 85% of all issuance, EUR denominated issuance accounts for over 13% with a smattering of issuance in GBP, JPY, AUD and CHF, as well. Expansion of issuance into non-USD denominations suggests the growing acceptance of this CDO sub sector in Europe and Asia, as well as the inclusion of non-USD denominated collateral in the CDO underlying portfolios. “IT’S WHAT’S INSIDE THAT COUNTS”
While it is not mere happenstance that the spurt in SF CDO issuance has coincided with a booming US housing market, that in itself does not explain the full story. In fact, early vintage SF CDOs (pre-2003 vintages) had a rather limited exposure to the US housing sector. They were multi-sector CDOs and included exposure to a wide variety of asset classes including some sectors that did not have an established track record such as manufactured housing, franchise loans and aircraft leases whose performance has continued to remain poor. The motivation behind the multi-sector CDOs was to achieve high diversity scores by constructing portfolios with exposure to a large number of asset classes. The diversification benefits hoped for did not materialize as some sectors underperformed resulting in downgrades of many SF CDOs. As a result, SF CDOs shifted their focus to collateral in mainstream ABS sectors such as RMBS, CMBS and consumer ABS with established track record. We illustrate the significant differences in the collateral composition of SF CDOs by examining a sample of 60 CDOs comprised of vintages from 2001-2005 (Annex 1), using data from Intex. This selection is substantial in size and a representative sample of SF CDOs issued over this period. For each year, the number of transactions in the sample is representative of the type of SF CDOs issued in that year. As such, for 20012003, the sample consisted entirely of mezzanine ABS CDOs, and for 2004-05, we included high grade ABS CDOs, in addition to the mezzanine deals.
1 2
In much of this report, we focus on mezzanine and high grade. We have included ABCP/money market tranches in our deal size computations.
Please see additional important disclosures at the end of this report.
359
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 54
For brevity of presentation, we have aggregated the Moody’s sector representation into a smaller number of aggregate sectors. Still, it is useful to point out that the average number of Moody’s sectors in the 2001-02 deals in our sample is over 18 and the same for 2003-05 deals is only about 11, which serves to illustrate the multi-sector nature of the pre-2003 transactions compared to the latter vintages, which have focused on fewer but better established sectors. EXPOSURE TO DISTRESSED ABS SECTORS
In our sample, exposure to distressed ABS sectors such as manufactured housing, franchise loans, aircraft leases, equipment loans and structured settlements averaged about 15% of the collateral pool in the SF CDOs of 2001 and 2002 vintages, ranging from a low of 3.7% to a high of 40.7%. For 2003-2005 vintages, exposure to the same distressed sectors averaged about 0.8%. As shown in Exhibit 2 below, these sectors have continued to perform poorly. Downgrade Rates in Distressed ABS Sectors
exhibit 2 ABS Sector Manufactured Housing Aircraft Leases Equipment Leases Franchise Loans Tobacco Settlements
Ratings Outstanding on 1/1/2004 671 62 151 156 60
Number of Downgrades in 2004 276 41 50 66 55
Downgrade Rate 2004 41% 66% 33% 42% 92%
Source: "Structured Finance Ratings Transitions: 1983-2004", February, 2005, Moody's
The link between the poor performance of SF CDOs and their exposure to the distressed ABS sectors has previously been documented (See “The Usual Suspects”, December 17, 2004). Even though we have not included SF CDOs from the 1999 and 2000 vintages in Annex 1, they also have levels of exposures comparable to the 2001-02 SF CDOs. As such, post-2003 vintages have a markedly different collateral composition compared to the previous vintages. This difference is reflected in the ratings performance of SF CDOs. To demonstrate this, we updated a study on rating actions we did earlier in the year (“Understanding Asymmetry in CDO Ratings Migration,” February 11, 2005). There were 223 downgrade actions pertaining to SF CDOs within the dataset. The distribution of these downgrade actions by vintage is shown below (Exhibit 3). All but two downgrades pertained to 1999-2002 vintages, which suggests a continuing strong link between exposure to the distressed ABS asset classes and the consequent CDO performance. Most recent transactions have explicit limits on exposures to these distressed ABS sectors – in particular, to manufactured housing. Moody’s SF CDO Downgrades by Vintage
exhibit 3 Vintage 1999 2000 2001 2002 2004 Grand Total Source: Moody's
360
2002 1 3
4
Year of Rating Action 2003 2004 12 2 27 56 5 31 11 44
100
2005 Grand Total 15 15 98 31 70 27 38 2 2 75 223
EXPOSURE TO RMBS
The shift towards increased concentration of the RMBS sector in SF CDO portfolios is evident from Exhibit 4. The average exposure of the 2003-05 cohort to the RMBS sector in our sample is nearly double the exposure for the 2001-02 cohort. While exposure to both prime and sub-prime RMBS has increased, in the 2003-05 cohort, the increase in exposure to sub-prime RMBS is significantly higher.3 We find some difference in RMBS exposure between mezzanine and high grade transactions in our sample. On average, mezzanine SF CDOs had about 6% lower exposure to prime RMBS and about 7% higher exposure to sub-prime RMBS relative to high-grade SF CDOs. exhibit 4 Cohort 2001-02 Mezzanine ABS 2003-05 Mezzanine ABS 2003-05 High grade ABS
SF CDO Exposure to the RMBS Sector Prime 10.30% 16.30% 22.30%
Sub-Prime 30.20% 56.90% 50.20%
Total 40.50% 73.20% 72.50%
Source: Morgan Stanley, Intex
The dominance of RMBS collateral, particularly sub-prime is not surprising. The continued strong performance to date of the RMBS sector – both prime and sub prime categories – is reflected in their ratings performance. Attractive spreads relative to comparably rated ABS securities, coupled with the robust new issuance volumes in RMBS, have buoyed the lure of RMBS to SF CDO managers. Even though the strong CDO bid has contributed to a significant tightening of RMBS spreads more recently, they have remained attractive relative to other ABS asset classes, particularly at the BBB level, with the possible exception of CDO/CBO tranches. In Exhibit 5, we compare the ratings performance of the US RMBS securities over the last 10 quarters by significant RMBS sub-sectors. While upgrades continue to be a multiple of downgrades both in the prime and sub-prime categories, ratings performance in terms of upgrades has been significantly stronger in the prime category. In all, the superior ratings performance of the RMBS sector across the board is the result of high prepayments driven in part by relatively low mortgage interest rates, seasoning of the underlying loans and housing market appreciation. HIGH RMBS CONCENTRATION AND CDO PERFORMANCE
It is logical to inquire into the performance consequences to date of SF CDOs with high concentrations of exposure to RMBS. Admittedly, the substantial pick-up in the SF CDO issuance volume was a second half of 2004 phenomenon and it is probably too soon to draw conclusions on their performance. Still, ratings stability to date of the 2003-05 cohort of SF CDOs provides an early comfort. As we have pointed out earlier, only two out of a total of 223 Moody’s downgrades of SF CDOs were from this cohort. We are also comforted by a recent S&P study 4 which points out that of the 12 SF CDO transactions upgraded (a combination of 2001-03 vintages), 10 transactions had exposure to RMBS in excess of 60% of the collateral portfolio. 3
For this report, we included jumbo loans, residential A and Alt A collateral under RMBS prime and residential B/C loans, home equity loans and HELOCs under RMBS-sub-prime. 4 “CDO Spot Light: Is the Fortune for Structured Finance CDOs Tied to RMBS Performance for Better or Worse”, Standard and Poor’s, September 2005.
Please see additional important disclosures at the end of this report.
361
362
exhibit 5 500 450 400 350 300 250 200 150 100 50 0 -50 2Q03
Upgrades/Downgrades
3Q03
4Q03
1Q04
Prime Jumbo
3Q04
HEL: Sub Prime
2Q04
Alt A
4Q04
Total RMBS
1Q05
2Q05
3Q05
chapter 54
Source: S&P
Number of Upgrade and Downgrades in US RMBS Sector
Structured Credit Insights – Instruments, Valuation and Strategies
SF CDO EXPOSURE TO CDO/CBO SECTOR
CDOs are one asset class that offer attractive spreads versus nearly all other ABS asset classes. Most SF CDOs have upper limits on exposure to other CDO/CBO assets. Still, inclusion of CDO collateral in SF CDO portfolios, to the extent permitted, has become attractive from the perspective of SF CDO managers. Exhibit 6 shows the average exposure of SF CDOs to the CDO/CBO sector. We aggregated exposures by the underlying CDO asset class, broadly into Corporate (including emerging market CDOs) and SF CDOs. exhibit 6 Cohort 2001-02 Mezzanine ABS
SF CDO Exposure CDO/CBO Sector Corp CDO 4.68%
SF CDO 4.27%
Total 8.95%
2003-05 Mezzanine ABS
3.87%
2.20%
6.08%
2003-05 High Grade ABS
3.67%
10.96%
14.63%
Source: Morgan Stanley, Intex
Three points are worth making in this context. First, there is a fair degree of variation within our sample to exposure to the CBO/CDO sector. As such, one should be careful with the use of averages. Second, we note that high yield CBOs account for the majority of the corporate CBO exposure in our sample. Further, investment grade tranches of CDOs/CBOs account for the bulk of the CDO/CBO exposure. Third, we are somewhat surprised by the average level of exposure to other SF CDOs within SF CDO portfolios. This has significant implications in terms of risk overlap and manager incentives and deserves a more detailed analysis. COLLATERAL COMPOSITION DIFFERENCES BETWEEN HIGH GRADE AND MEZZANINE SF CDOS
In Annex 2, we compare ratings distributions of the underlying collateral between high grade and mezzanine SF CDOs. We limit this comparison to 2004 and 2005 vintages for which our sample includes both mezzanine and high grade transactions. By definition, high grade SF CDOs have highly rated collateral relative to the mezzanine SF CDOs. It is interesting to note the significant variation within both mezzanine and high grade SF CDO portfolios in terms of ratings composition. Within high grade transactions, there are a few with high concentration towards the A2/A3 range. Within mezzanine transactions, Baa2 and Baa3 ratings buckets clearly dominate the collateral composition. STRUCTURAL INNOVATIONS
SF CDO market has continuously evolved, adapting to change as market conditions have changed and introducing innovations. We have already discussed the changes in their collateral composition as a reaction to the performance of the underlying collateral sectors. These innovations include the growing applications of synthetics, below investment grade buckets, pro-rata paydowns and senior funding through term notes. In this section, we discuss recent structural innovations in SF CDOs.
Please see additional important disclosures at the end of this report.
363
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 54
WHY SYNTHETICS MATTER TO CASH SF CDOS?
Synthetics in SF securities in the context of SF CDOs and in single name contracts referencing SF securities are not entirely new. In some form, they have been around for at least three years. However, the absence of a “street-standard” documentation had clearly hampered their widespread application and limited their liquidity. The publication of ISDA standard confirmation in late June 2005 has cleared the way for the standardization of contract terms and documentation (See “Understanding Synthetic Structured Finance – First Steps”, July 12, 2005 for a detailed discussion on the topic). Synthetics have significant applications to SF CDOs in three important ways. Efficient Funding: Synthetics enable SF CDOs to achieve significant funding efficiencies, particularly at the senior/super senior parts of the capital structure, amounting to cheaper funding by as much as 15 bp depending upon the structure and the credit quality of the underlying collateral. As a result, a number of mezzanine SF CDO transactions no longer use money market classes in their capital structure. Collateral Sourcing: Difficulties in sourcing collateral have been a significant constraint facing mezzanine SF CDO managers. By their very nature, SF securities have a smaller supply of BBB tranches relative to tranches rated higher. Difficulties in sourcing collateral meant long ramp-up periods and potential for negative carry. Through the application of synthetics, such constraints are relieved significantly as the availability of physical cash bonds is not a binding constraint. Security Selection: The availability of collateral in the cash SF market is largely in the new issue market and some managers may prefer older vintage deals or certain issuers. For instance, some investors may be uncomfortable with the current new issue deals, particularly in the sub-prime HEL securities because of their higher exposure to affordability products. Synthetics also enable managers to reference older vintages of HEL securities, which may have lower exposure to affordability mortgage products. Synthetics significantly expand the universe of collateral a manager can access which should enable managers to distinguish themselves instead of being constrained to having to invest in what is available in the new issue market. They also enable managers to achieve vintage diversification within CDO portfolios, which might be a key determinant of performance going forward. GROWING SYNTHETIC BUCKETS
The advantages offered by the use of synthetics have resulted in growing synthetic buckets in SF CDOs, both in terms of their frequency and size. We examined 140 new issue reports of SF CDOs issued between 2004 and 2005. The proportion of transactions with an explicit synthetic bucket rose from 24% in 2004 to 52% in 2005. The average size of synthetic buckets across all transactions doubled between 2004 and 2005. The largest synthetic bucket in our sample was 25% in 2004 deals and the same was 65% in 2005 deals. These statistics point to the growing recognition of the advantages offered by synthetics. HYBRIDS ARE COMING
With increasingly larger synthetic buckets, the distinction between cash and synthetic CDOs has begun to fade. Hybrid structures provide the flexibility for managers to choose between acquiring risk exposure to assets in cash or synthetic form and pick
364
assets where best execution is possible regardless of form. Not only does this imply shorter ramp-up periods but also the retention of the many structural features of typical cash CDOs (such as the coverage and credit quality tests, cash flow diversion triggers and the like). Further more, hybrid structures can also achieve funding efficiencies by the use of unfunded super senior tranches. In our opinion, this is the direction the SF CDO market is headed. BIG IS GETTING BIGGER
As BBB spreads have narrowed, SF CDOs striving for yield have pushed for provisions to allocate a portion of the collateral to below investment grade (BIG) assets to achieve targeted equity returns. Examining 2005 transactions, we found BIG buckets in over half of the transactions, which averaged about 10%, with a few transactions that allowed greater than 30% allocation for BIG assets. PRO-RATA PAYDOWN
Several recent deals include pro-rata paydown of the principal waterfall. This feature originated in high-grade deals and now spawns mezzanine deals as well. A pro-rata paydown is designed to allow all classes of debt in the capital structure to be paid down in proportion to their outstanding balances as long as there has not been any breach of a coverage test and a certain proportion of the collateral balance is still outstanding. If there is a breach of a coverage test, pro-rata paydowns will stop and cash flows revert to the traditional CDO waterfall mechanism. The purpose of this feature is to prevent the cost of funds to the deal from increasing as transactions de-lever and senior notes, which have the lower cost of funding, are paid off. This results in maintaining equity returns as well as shortening the weighted average lives of the mezzanine tranches. SENIOR FUNDING THROUGH TERM NOTES
ABCP tranches evolved as a hybrid of traditional cash CDOs and ABCP conduits. They started to become widely used in SF CDOs about two years ago as a means of reducing cost of funding for CDOs. CP tranches generally included third-party liquidity facilities for 100% of the principal amount of CP classes. CP tranche investors are repaid for the principal portion of the CP notes when they mature from the proceeds of new issuance of CP notes. If such new issuance or re-marketings fail, the liquidity facility providers will be relied upon for the repayment of CP notes as long as the credit quality of the collateral pool is performing within expectations. The relative advantage of having CP tranches has declined recently. We have noted several high-grade transactions that instead rely on term notes which do not carry the uncertainties associated with ultimate funding costs and the success of re-marketing.
Please see additional important disclosures at the end of this report.
365
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 54
RISKS
Despite the impressive growth and the constant innovation evident in the SF CDO market, a few risks inherent in the product concern us. We are not alone in being cautious about the US housing sector, their potential links with consumer spending and implications for the RMBS sector. Much has been written about this topic (for example, see Richard Berner’s report “Housing Wealth and Consumer Spending”, October 7, 2005). Given the exposure of SF CDOs, both high grade and mezzanine CDOs, to RMBS, the consequences are obvious. Credit losses on the collateral underlying RMBS will likely increase from their current low levels, if house price growth slows. Clearly, these risks are not unknown to market participants. We take some comfort that CDO structures and rating agency models require stress analysis that takes into consideration the prospect of higher credit losses in the underlying collateral. We are yet to be convinced that there is a nationwide housing bubble. Still, whether or not the US housing market avoids an unprecedented nationwide decline, concentrated exposure to one asset class does give us pause. We think that managers employ diverse strategies for mitigating risks inherent in RMBS, particularly the available funds cap (AFC) risk. In these tough-to-tread markets, manager expertise will be critical in determining which strategies succeed. We agree with the sentiment that manager selection matters more today than ever. Investors should understand CDO managers’ approach towards dealing with the risks in RMBS across different credit cycles. We understand the diversity benefits of including CDO/CBO collateral within the SF CDO portfolios. Yet, that diversity might not come cheap. Corporate CDO assets in SF CDO portfolios seem like off-market bets to us. There are issues associated with issuer overlap such exposures bring that investors should watch for. Similarly, large below investment grade buckets need to be understood in an appropriate risk-return context. While we generally welcome synthetics to the SF CDO sector, it is important that investors be aware that they bring with them several operational and documentation challenges, as well as counterparty risk exposure issues. Cash bonds and synthetics that reference them differ significantly in their exposure to AFC risk. Investors should understand the nature and mechanics of synthetics and satisfy themselves that CDO managers have the necessary systems in place to deal with the challenges of synthetics.
366
Please see additional important disclosures at the end of this report.
367
2002 2002 2002 2002 2002 2002 2002 2002 2002
Saybrook Point CBO II
Duke Funding III
Fort Point CDO I Limited
Independence CDO III
Newcastle CDO -I
South Coast Funding II Ltd.
ABS Capital Funding II
Fulton Street CDO
Solstice ABS II
Mezzanine ABS
2001
2001
Putnam Structured Product CDO 20001-1 2002
2001
MKP CBO II
South Coast Funding I
2001
Independence CDO II
ACA ABS CDO 2002-1
Mezzanine ABS
2001
Duke Funding II
Mezzanine ABS
Mezzanine ABS
Mezzanine ABS
Mezzanine ABS
Mezzanine ABS
Mezzanine ABS
Mezzanine ABS
Mezzanine ABS
Mezzanine ABS
Mezzanine ABS
Mezzanine ABS
Mezzanine ABS
Mezzanine ABS
Mezzanine ABS
2001
Diversified Asset Securitization Holdings III
SF CDO Type Mezzanine ABS
Vintage 2001
4.6
5.39
11.25
19.2
0
11.63
10.88
7.72
12.59
11.37
20.99
10.29
11.67
14.55
9.96
12
RMBS Prime 1.31
20.37
17.31
41.28
52.29
3.27
18.99
47.93
49.85
42.58
45.77
51.34
19.29
21.84
11.8
18.99
34.68
RMBS Sub-Prime 15.19
3.65
21.6
15.15
6.86
6.31
16.01
3.87
11.59
8.09
13.44
5.93
11.53
19.24
23.27
21.07
20.4
Distressed ABS 40.72
3.15
9.57
0.59
9.59
80.24
31.5
22.47
20.99
11.86
12.42
8.67
34.62
10.34
43.76
43.08
17.5
CMBS/ REIT 16.33
27.9
0.59
8.09
5.78
0
3.66
7.41
0
5.76
0
6.54
1.28
6.9
4.57
0
0
Corp/ EM CDO 1.08
32.25
13.35
1.35
0
4.97
13.64
0
0
0
4.57
0
0
0
0
0
0
SF CDO 2.5
Collateral Composition of Select SF CDOs by Vintage and Underlying Collateral (% of Composition)
Name Arroyo CDO I
annex 1
8.07
32.2
22.3
6.28
5.21
4.58
7.45
9.86
19.14
12.42
6.51
22.95
30
2.04
6.9
15.44
Other 22.85
Name ACA ABS 2003-1 Ltd. C-BASS CBO VII Duke Funding V Fort Point CDO II Ltd. Independence CDO IV Solstice III South Coast Funding IV South Coast Funding III ACA ABS 2003-2 Saturn Ventures SF CDO ACA ABS 2004-1 Crystal Cove Knollwood Straits Global Duke Funding VI Dunhill ABS CDO Alexander Park Glacier CDO Independence V CDO Inman Square Funding Ischus ABS CDO I Newcastle V South Coast Funding V Vermeer Funding II
annex 1 (cont.)
Vintage 2003 2003 2003 2003 2003 2003 2003 2003 2003 2003 2004 2004 2004 2004 2004 2004 2004 2004 2004 2004 2004 2004 2004 2004
368 RMBS Prime 10.97 22.51 26.12 12.72 12.31 3.77 18.51 14.96 12.2 10.51 15.81 23.63 17.31 20.16 27.06 15.08 26.15 33.38 26.38 19.38 23.03 0 13.12 18.2
RMBS Sub-Prime 53.35 46.44 61.16 57.13 53.73 65.26 60.79 58.56 60.5 52.8 60.6 44.15 61.13 60.96 45.56 71.48 55.88 47.3 55.45 74.94 62.54 22.6 50.22 59.99
Distressed ABS 3.97 4.49 0 0 5.7 0 0 3.32 0 0 0 6.78 0 0 0 0 0 0 0 0 0 0 0 0
CMBS/ REIT 5.32 0 1.39 19.86 11.19 2.7 11.1 13.78 13.12 21.27 14.69 8.64 12.9 3.29 20.45 4.78 12.8 11.48 8.74 0 8 72.75 26.55 6.48
Corp/ EM CDO 8.77 0 0 4.57 1.26 12.29 3.99 5.46 1.57 3.66 5.06 1.11 5.41 4.62 0 4.45 0 0 5.11 0 5.63 0 2.02 6.71
SF CDO 0.55 6.07 0 1.01 5.47 7.2 0 0 3.42 3.67 0 0.33 0 3.85 0 3 0 6.79 0 2.8 0 0 2.2 1.98
Other 17.09 20.49 11.33 4.73 10.32 8.8 5.6 3.91 9.17 8.11 3.84 15.38 2.92 7.13 6.93 1.21 5.17 1.05 4.32 2.88 0.81 4.66 5.89 6.62
chapter 54
SF CDO Type Mezzanine ABS Mezzanine ABS Mezzanine ABS Mezzanine ABS Mezzanine ABS Mezzanine ABS Mezzanine ABS Mezzanine ABS Mezzanine ABS Mezzanine ABS Mezzanine ABS Mezzanine ABS Mezzanine ABS Mezzanine ABS Mezzanine ABS Mezzanine ABS Mezzanine ABS Mezzanine ABS Mezzanine ABS Mezzanine ABS Mezzanine ABS Mezzanine ABS Mezzanine ABS Mezzanine ABS
Collateral Composition of Select SF CDOs by Vintage and Underlying Collateral (% of Composition)
Structured Credit Insights – Instruments, Valuation and Strategies
Please see additional important disclosures at the end of this report.
369
2005 2005
Kent Funding Ltd.
Klio II Funding
Source: Morgan Stanley, Intex
2005
High Grade Structured Credit CDO 2005-1 Ltd.
High Grade ABS
2004 2005
2004
Pinnacle PT Funding
Zenith Funding Ltd
2004
Davis Square III
Athos Funding Ltd. 2005-1
High Grade ABS
2004
Streeterville ABS CDO
High Grade ABS
High Grade ABS
High Grade ABS
High Grade ABS
High Grade ABS
High Grade ABS
High Grade ABS
High Grade ABS
2004 2004
Sierra Madre Funding
High Grade ABS
High Grade ABS
Mezzanine ABS
Mezzanine ABS
Mezzanine ABS
Duke Funding High Grade I
2004
Davis Square II
High Grade ABS
2004
Duke Funding High Grade I 2004
2005 2004
Camber 3
Blue Heron IX
2005
Saturn Ventures 2005-1
Cascade Funding I
High Grade ABS
2005
Orchid CDO Limited II
Mezzanine ABS
2005
Commodore CDO III
SF CDO Type Mezzanine ABS
Vintage 2005
9.05
18.49
17.31
15.43
18.98
17.75
23.59
34.56
28.1
34.83
23.84
21.06
14.21
34.83
6.71
12.35
10.36
2.08
RMBS Prime 19.08
52.09
57.24
72.99
58.76
55.32
45.32
52.06
47.03
48.72
42.12
34.96
54.75
39.6
42.12
64.36
60.75
53.7
70.24
RMBS Sub-Prime 58.8
0
0
0
0
0
0
0
0
4.56
0
0.14
0
0
0
0
5.3
0
0
Distressed ABS 0
0.63
5.6
0
1.07
6.62
3.32
15.81
1.58
14.63
11.8
6.65
0
29.53
11.8
13.02
4.07
8.35
7.34
CMBS/ REIT 11.16
16.98
3.21
1.69
0
5.04
4.14
8.52
0
2.39
0
8.58
0
0.83
0
0.77
0
22.86
4.25
Corp/ EM CDO 2.73
18.14
15.26
8.02
3.59
12.83
21.43
0
16.62
0
0.08
23.94
24.19
9.35
0
0
5.08
0
3.53
SF CDO 6.96
Collateral Composition of Select SF CDOs by Vintage and Underlying Collateral (% of Composition)
Name ACA ABS 2005-1
annex 1 (cont.)
1.61
0.2
0
21.16
1.21
8.04
0
0.21
1.32
11.24
1.87
0
6.47
11.24
15.14
12.45
4.73
12.57
Other 1.27
370 2004 2004 2004 2004 2004
Ischus ABS CDO I
Newcastle V
River North CDO
South Coast Funding V
Vermeer Funding II
2005
2004
Inman Square Funding
Camber 3
2004
Independence V CDO
2005
2004
Glacier CDO
Saturn Ventures 2005-1
2004
Alexandre Park
2005
2004
Dunhill ABS CDO
Orchid CDO Limited II
2004
Duke Funding VI
2005
2004
Straits Global
2005
2004
Knollwood
Commodore CDO III
2004
Crystal Cove
ACA ABS 2005-1
2004
ACA ABS 2004-1
Mezzanine ABS
Mezzanine ABS
Mezzanine ABS
Mezzanine ABS
Mezzanine ABS
Mezzanine ABS
Mezzanine ABS
Mezzanine ABS
Mezzanine ABS
Mezzanine ABS
Mezzanine ABS
Mezzanine ABS
Mezzanine ABS
Mezzanine ABS
Mezzanine ABS
Mezzanine ABS
Mezzanine ABS
Mezzanine ABS
Mezzanine ABS
Mezzanine ABS
Vintage SF CDO Type
0
0.79
1.99
1.11
4.56
1.51
13.95
3.95
0
0
0
0.18
5.29
7.33
0.65
22.57
0
4.95
9.78
10.67
Aaa
0.68
1.17
0.92
0.69
6.97
2.24
2.03
Aa2
0
1.32
0
0
0
0
0.63
0
0
0
0
0.21
1
0
3.17
0
5.23
0
0.76
1.33
0
0
0
0.4
0.7 11.05
3.22
0.4
0
0.77
1.79
0
0
Aa1
0
0.75
0
0
0
0
0
0.86
0
0
0
0
1.58
1.69
0.67
1.25
0
0.67
0
0.9
Aa3 8.55
6.98
A2
5.85
1.4
2.88
5.92
13.5
3.67
11.2
6.31
0
4.39
1.67
0
9.28 37.63 30.24
1.35 18.62 35.92
8.06 24.51 24.62 2.96 16.51 33.67 34.25
7.18 13.67 33.03 23.15
6.11
4.02 11.05 35.32 23.07
7.33
5.22
7.12
0
6.49
0
0
0
0
0
5.29
0
Ba1
0
0
0
0
0
0
0
3.2
0
Ba2
3.72 21.88 46.01 17.61
8.97 13.16 32.92 25.26
4.06
4.16
1
0
0.78
1.34
0
0
0.8
0
0.5
0.52
0
0
0
0
0
0.67
8.34
0
7.43 19.84 50.23 12.81
7.1 28.34 44.89 16.41
0
18
31 10.99
4.37 26.07 44.72
7.93 10.11
6.72 19.71 38.19 10.81
6.66 10.56 19.31 42.05 13.59
0.72 10.78
1.21 10.57
0
0
0.39
3.99
7.12 22.87 35.96 17.77
1.04 10.49 10.69
0
0
0.13
1.36
0 11.65
1.14 12.25
9.59
1.54 27.56 42.33 20.62
3.76 10.59 32.88 14.72
3.73 15.28 27.51 12.25
2.72 11.59 40.16 24.93
A3 Baa1 Baa2 Baa3
3.25 14.12 20.91 23.41
0
0.54 22.44
0.86
0
A1
0
0
0
0
0
0
0
0
0.62
0
1.32
0
0
0
0
0
0
0
0
0
Ba3
0
0
0
0
0
0
0
0
0
0
0
B1
0
0
0
0
0
0
0
0
0.83
Ratings Distribution of Mezzanine and High Grade CDO Portfolios
0
0
0
0
0
0
0
0
1.01
0
5.49
0
0
0
0
0
0
0
0
0
B2
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
NR/ WR
1.85
0
0
0
0
0
0
0.64
0.67
0
2
0
0
0
0
0
0
0.64
0
0
0.62
0 30.17
0
0
0 10.29
0
0
0
0
0
0
0
0
0
0 11.32
0
Ca & B3 Below
chapter 54
Name
annex 2
Structured Credit Insights – Instruments, Valuation and Strategies
Please see additional important disclosures at the end of this report.
371
High Grade ABS High Grade ABS
2004 2004
Cascade Funding I
Davis Square II High Grade ABS
High Grade ABS High Grade ABS
2004 2004 2004 2004 2004
Sierra Madre Funding
Streeterville ABS CDO
Davis Square III
Pinnacle PT Funding
Zenith Funding Ltd
Athos Funding Ltd. 2005-1 2005
2005 2005
Kleros Preferred Funding
Klio II Funding
Source: Morgan Stanley, Intex
2005
Kent Funding Ltd. High Grade ABS
High Grade ABS
High Grade ABS High Grade ABS
High Grade Structured Credit CDO 2005-1 Ltd. 2005
High Grade ABS
High Grade ABS
High Grade ABS
High Grade ABS
Duke Funding High Grade I 2004
High Grade ABS
High Grade ABS
2004
Blue Heron IX
Vintage SF CDO Type
Duke Funding High Grade I 2004
Name
annex 2 (cont.)
32.4
Aa2
9.81
32.4
1.88 36.06
7.89 28.96
5 18.08
9.81
Aa1
33.6
5.03 14.75 6.47
42.98 16.88 19.25
A1
1.72
9.32
2.06
1.53
3.77
9.55
9.17
1.92
4.56
11.3
7.97
0.4
0.18
2.95
3.64
5.47
8.99
2.58 15.24
34.2
16.8
7.22
0.35 12.92
6.65
2.76 14.91
6.11 12.57
6.4 19.36
1.14
5.32
4.56
A2
0 17.47
3.09
9.2 18.58
3.85
3.36
4.88
9.2 18.58
Aa3
39 16.67
5.41 28.55
1.14
4.23 53.33
6.07 26.58
5.4 31.93
21.64 26.04 17.24
35.35
1.6
40.08
26.65
26.28
37.68
41.89
43.28 12.18 18.92
23.04
33.39
34.26
62.6
23.04
Aaa
0.61
8.1
4.14
9.8
6.78
9.13
0.56
5.67
0
2.21
1
3.84
8.06
1.03
1
0
0
0.5
3.78
0
0
0
0.07
0
1.41
0
0
0
0
0
0
0
0.19
0
0
0
0.61
0
0
0.41
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0.04
0
0
0
0
0
A3 Baa1 Baa2 Baa3
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
Ba1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
Ba2
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
Ba3
Ratings Distribution of Mezzanine and High Grade CDO Portfolios
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
B1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
B2
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
Ca & B3 Below
2.29
5.19
0
5.09
0.84
0
0
0
1.03
1.29
1.41
5.91
0
0
1.41
NR/ WR
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 55
Are Senior Tranches Sticky?
November 10, 2005
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj We reflect on how the concept of convergence may or may not apply to today’s credit markets. At one level, we argue that convergence has already happened, at least within the space of pure corporate credit instruments like bonds, loans and derivatives. Yet there are still plenty of opportunities for convergence at both ends of the macro/micro spectrum, from corporate finance themes impacting equity and credit markets simultaneously to cash and synthetic CDO tranches pricing similarly. One potential convergence theme we focus on in this chapter is more micro than macro, and is related to the relationship of highly rated tranches across the corporate and securitized credit markets. Senior tranches have experienced shifts in valuation as markets matured, most notably in structured mortgage products (CMOs) and in ABS and CMBS in the mid 1990s, and perhaps also in cash CLOs more recently (see Exhibit 5). In some respects, this phenomenon can be thought of as a complexity premium for new products that have not fully matured in terms of market presence. Therefore, newer instruments should tend to price cheaper than established instruments with roughly equivalent risks (and justifiably so).
372
During the last credit cycle, senior synthetic CDO tranches traded much like the rest of the capital structure, in that there was generally no bid, and forced sellers (because of ratings downgrades) were met with a market that had little appetite to absorb the risk. These tranches were eventually sold into the hands of investors who believed that they were cheap, given the low likelihood of default, and that they would rally back substantially as the credit cycle moved forward. They were right (see Exhibit 1). As we see it, there are three dimensions to risk in senior tranches, namely the likelihood of principal losses, ratings sensitivity, and mark-to-market issues related to leverage and negative convexity. How will such tranches behave in a negative credit environment? The jury is still out on whether we have seen a true secular shift in the pricing of senior tranches. However, we believe that such tranches could exhibit substantially less price sensitivity than implied by correlation models under certain market conditions. Supporting factors include the strong presence of senior note buyers in the market who look across credit markets for opportunities, and Basel II factors that are very supportive of taking credit risk in highly rated instruments. But not all senior tranches are created equal, and there are certainly many such tranches that are weak and as such may have much more ratings sensitivity than others. There is also the possibility that a more systemic event makes these tranches even more sensitive than the models predict. exhibit 1
Repricing AAA Synthetic Tranches – Cyclical or Secular? (CDX 5 Yr)
180
Index 7-10% Mid
150
120
90
60
30
0 Oct-03
Feb-04
Jun-04
Oct-04
Feb-05
Jun-05
Oct-05
Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
373
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 55
RISK #1: HOW LIKELY ARE PRINCIPAL LOSSES?
Typical attachment points for senior tranches are well above average loss rates for investment grade portfolios, which range from 1.2% over 5-years to 2.4% over 10years, for a pure Baa-rated portfolio (Moody’s). Yet, there are two problems with such a simple analysis. First, default rates are volatile and very much cyclical, and second default risk in typically-sized portfolios (100-200 names) can be quite different than for the market at large (assuming the same average characteristics) due mainly to sampling risk. As we have addressed many times before in our research, portfolio sampling risk was a key driver of underperformance of underlying portfolios in CDOs relative to the market during the last credit cycle (see Chapter 32). If we take these points into consideration, the tail risks in losses for practical-sized portfolios increase (see Exhibit 2), and these results have meaningful implications on equity and mezzanine tranches in investment grade portfolios, but typical senior tranches still appear safe from principal losses. We should note that this phenomenon is very much an investment grade one. In the high yield market, typical portfolio size losses have a much greater probability of touching senior tranches (albeit still small). The implication is that high yield losses could penetrate senior tranches, especially given the vagaries of an unreceptive new issue market, which is a lesson we learned during the telecom bust. This also helps explain why senior notes on high yield collateral traded the way they did in the wake of that market environment as well as why there remains a basis between equivalent rated high yield and investment grade tranches today. exhibit 2
Distribution of 5-Year Losses – Simulation Based on Moody’s Data
45% Portfolio (125 Names) 40% Universe (1,000 Names) 35% 30% 25% 20% 15% 10% 5%
Source: Morgan Stanley
374
M or e
3. 75 %
or
75 %
25 %
3. 25 -3 .
75 %
2. 75 -3 .
25 %
2. 25 -2 .
75 %
1. 75 -2 .
25 %
1. 25 -1 .
75 %
0. 75 -1 .
0. 25 -0 .
0. 00 -0 .
25 %
0%
RISK #2: RATINGS SENSITIVITY
If the likelihood of principal losses is not a big risk factor in senior tranches, such tranches are not necessarily attractive investments at any price. Many who invest are doing so for ratings purposes, and as such are very sensitive to potential ratings downgrades. As we mentioned earlier, significant ratings downgrades did bring in a reasonable amount of sellers into the market in 2002 and early 2003. In a broad-brush attempt to get some sense of ratings sensitivity, we used S&P’s publicly available ratings model to estimate the ratings sensitivity of tranches (over a 2 year period) with certain attachment points (assuming the CDX 5 portfolio for 5, 7 and 10-year terms). The attachment points we considered are not uncommon for AAA and AA tranches, although there is a lot of variation (in both directions) in the market. A few scenarios can give us some insight (see Exhibit 3). Our analysis shows that downgrading the 15 worst credits (out of 125) by one notch could keep ratings on both AA and AAAs stable, and even at two notches for this group, only one of our example tranches realizes a downgrade. However, combining two-notch downgrades with three actual defaults results in tranche downgrades of one to four notches. Downgrades to all names in the portfolio are also material especially if combined with some defaults. exhibit 3
AA Tranche – 2 Years Forward Typical Attachment
How Sensitive Are Ratings – 2 Years Forward (CDX 5) 5 year 5%
7 year 6%
10 year 8%
15 Worst Dngrd 1 Notch
AA
AA
AA
15 Worst Dngrd 2 Notches
AA
AA
AA
A-
A
A+
15 Worst Dngrd 2 Notches + 3 Defaults 125 Dngrd 1 Notch 125 Dngrd 1 Notch + 3 Defaults AAA Tranche – 2 Years Forward Typical Attachment
AA
AA-
A+
BBB
BBB
A-
5 year 6%
7 year 7%
10 year 9%
15 Worst Dngrd 1 Notch
AAA
AAA
AAA
15 Worst Dngrd 2 Notches
AAA
AA+
AAA
15 Worst Dngrd 2 Notches + 3 Defaults
AA+
AA
AA
125 Dngrd 1 Notch
AAA
AA
AA
AA
A
AA-
125 Dngrd 1 Notch + 3 Defaults
Note: This analysis was performed using S&P’s CDO Evaluator 2.4.3. Certain assumptions are implicit in the model, including S&P’s assumed recovery rates, default rates and correlations between sectors and regions. Our estimations are based on these assumptions and are not to be construed as potential ratings actions. Source: Morgan Stanley, S&P
Please see additional important disclosures at the end of this report.
375
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 55
RISK #3: LEVERAGE, CONVEXITY AND MTM ISSUES
Clearly, ratings shifts are a risk senior tranche investors need to consider, and it probably deserves more attention than the actual probability of loss. But perhaps the biggest unknown going forward involves understanding how such tranches might trade in a broadly negative credit environment. The appetite for senior tranches during the last credit cycle was poor, although many in the market today do not benefit from having that experience. The 2003 credit rally along with the liquidity and transparency introduced by the index tranche market forced an important repricing of such tranches (see Exhibit 1), to a point where AAA versions trade comfortably through the underlying index. In the world of highly rated synthetic tranches, risk-neutral (i.e., correlation) models tell us that most senior tranches (excluding super seniors), are both levered to the underlying portfolio (delta > 1) and have negative convexity (deltas rise as spreads on the underlying rise). During this repricing, the deltas of these tranches fell dramatically (as they should), but when markets widened out this spring, the deltas continued their fall as senior tranches dismissed the event as idiosyncratic (see Exhibit 4). What happened to the negative convexity and will it ever return? exhibit 4
Where is the Negative Convexity? Deltas Once Fell as Spreads Tightened
Delta (x)
Spread (bp)
6
90 80
5 70 4
60 50
3 40 30 20
2 Index
Delta 1
10 0 Oct-03
Source: Morgan Stanley
376
Feb-04
Jun-04
Oct-04
Feb-05
Jun-05
0 Oct-05
LOOKING OUTSIDE OF CORPORATE CREDIT
A related issue may be the way the rest of the world looks if the corporate credit market is trading much wider than it is today. It is unlikely that corporate spreads move dramatically wider in isolation without risk premiums in other parts of the market moving as well. The extent to which senior synthetic tranches widen with the corporate credit market may be affected by the spread levels of similarly rated assets backed by different asset class (i.e. ABS, CMBS, RMBS and even cash CDOs) exhibit 5
Repricing AAs – CMBS and CLOs
bp
CMBS AA (Libor)
160
5 Year Swap Spreads CLO AA (Libor)
140 120 100 80 60 40 20 0 Dec-93
Dec-95
Dec-97
Nov-99
Nov-01
Nov-03
Oct-05
Source: Morgan Stanley
To the extent a widening of spreads is related to issues specific to corporate America (like increases in leverage) senior notes may have muted moves relative to what models may be pricing it. On the flip side, if widening spreads are driven by more macroeconomic factors (like a recessionary environment or a global dearth of risk capital ) then senior note pricing could move even more than today’s models may predict. Consider the 1998 period in Exhibit 5 as an illustration. Furthermore, portfolios of such tranches could come under pressure given the small universe of credits and overlap among them.
Please see additional important disclosures at the end of this report.
377
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 56
Taking a CLOser Look
November 21, 2005
Primary Analyst: Vishwanath Tirupattur Primary Analyst: Sivan Mahadevan The widely telegraphed troubles of the U.S. auto and auto parts sector, near simultaneous bankruptcy filings by Northwest and Delta and the much-ballyhooed pick-up in LBO activity suggest that the corporate credit environment is precariously perched. In fact, a generally negative sentiment seems pervasive across investment grade and high yield corporate bond investors. In contrast, the bull market in leveraged loans has continued unabated through much of the year, notwithstanding the troubles of the unsecured credit markets with healthy issuance and continuing tight spreads. The record issuance of CLOs is often credited (or blamed, depending upon one’s point of view) for the thriving leveraged loan market. A recent S&P report noted that inflows into the leveraged loan market over the past 24 months have exceeded the available supply by as much as $8.9 billion. 1 The increasingly gloomy corporate bond markets stand in a dichotomous contrast with the exuberance in the leveraged loans and their securitized product cousins, CLOs. On the other hand, this continued exuberance is not devoid of risks – leverage creep and declining credit quality, for example. Further, there is a growing trend in CLOs towards larger buckets of second lien loans, bonds and DIP loans. As we have pointed out in some of our previous reports (see Chapter 50 and “Tumult in Corporate Credit: Implications for Cash CDOs”, April 15, 2005), some of these provisions bring additional dimensions of risk-return calculus to CLO investments with a potential to affect different tranches in a very different manner. Given this backdrop, we think it is useful to understand the composition of CLO portfolios in terms of credit quality and sector composition to evaluate where CLO portfolio managers are placing their bets, which could be juxtaposed against the sector recommendations of our high yield strategists. Further, given the constrained supply of new issues and limited opportunities to source attractively priced loans in the secondary market, it is useful to understand the degree of overlap in terms of sectors and issuers across CLO portfolios. An analysis of the composition of CLO portfolios also helps appreciate the degree to which CLO portfolio managers are utilizing some of the recent provisions in terms of larger buckets of unsecured loans, second lien loans, DIP loans and bonds. In light of the LBO concerns, it is also useful to get a sense of the extent to which CLOs are exposed to recent LBO activity.
1
378
“Leveraged Lending Review 3Q05”, Standard and Poor’s, October 2005.
BOOMING CLO ISSUANCE
Global CLO issuance has been booming, with issuance volume year-to-date amounting to over $60 billion, about two times the entire 2004 issuance. Growth in new issuance has been impressive on both sides of the Atlantic (Exhibit 1) and has been at a sustained pace throughout the year. exhibit 1
Growth in Global CLO Issuance
Bn 70 Other 60
Euro USD
50 40 30 20 10 0 1999
2000
2001
2002
2003
2004
2005 TYD
Source: Morgan Stanley, Moody’s, S&P, Fitch IBCA, MCM Corporate Watch, Asset-Backed Alert, Intex
The strong issuance is reflective of several factors. First, ratings performance and stability of CLO tranches continue to be impressive. Over the 18 months, there have been 53 upgrades and 5 downgrades of CLO tranches, a very healthy ratio of over 10 to1. Even the five downgrades pertain to vintages of 2000 and prior. Ratings of the vast majority of outstanding CLO tranches were unchanged. Second, the new regulatory capital regime under Basel II largely favors securitized products thus contributing to the growing global acceptance of the CLO product. Further, following a difficult year, faced with deteriorating credit fundamentals, rising rates and a flat yield curve, there is an increasing preference by high yield investors in favor of secured loans versus unsecured bonds. High loan recoveries upon default remain an attractive story. Loans of recent high profile defaulting issuers such as Delphi continue to trade close to or even above par.
Please see additional important disclosures at the end of this report.
379
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 56
LEVERAGED LOANS AND CLO SPREADS
Despite a slight widening over the last few weeks, CLO spreads remain tight. Leveraged loan spreads also remain tight, even though they are somewhat wider than the levels seen earlier during the year. Historically, there is a strong relationship between leveraged loan spreads and CLO funding costs as depicted in Exhibit 2. exhibit 2
CLO Funding Costs and Leveraged Loan Spreads
Spread (bp)
Funding Cost (bp)
450
100 Funding Cost (LA) B+/B (RA)
90
400
BB/BB- (RA) 80 350 70 300 60 250 50 200
40 30 Feb-02
Oct-02
Jul-03
Apr-04
Jan-05
150 Oct-05
Source: Morgan Stanley
There is a statistically significant and strong positive correlation between leveraged loan spreads and CLO funding costs. Based on monthly data from 2002, we estimate the correlation coefficients between CLO funding costs and B+/B loans spreads and BB/BB- loan spreads to be 0.74 and 0.77, respectively. While we did not find statistically significant lead-lag relationships between leveraged loan spreads and CLO funding costs in either direction, it is clear that the two markets exert significant influence on each other. Exhibit 3 shows a time series of the share of different investor groups in the primary institutional markets over a longer period. It is clear that over the last few years, CLOs continue to be the largest investors of leveraged loans with a share of over 60% over the last few years. exhibit 3
Investor Share of Primary Leveraged Loan Market
100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 YTD
CLO
Source: S&P LCD
380
Insurance
Prime Rate Fund
Hedge Funds
Finance Co.
DATA
Using Intex data, we examined the portfolio composition of 40 U.S. CLOs issued between 2004 and the first half of 2005 with assets totaling $16.6 billion. Our sample of transactions covers about half the U.S. CLOs issued during 2004 and 30% of CLOs issued during the first half of 2005 and includes 31 managers. Given the relatively small overlap among managers, we think that our sample covers a diverse set of managers with a variety of management styles and provides a reasonable basis for understanding CLO portfolio composition. CLO PORTFOLIOS: BETTING THE SAME WAY?
The collateral composition by sector of each of the 40 CLO portfolios included in our sample is detailed an Annex 1. Industrials, services including retail, media, leisure, health care and telecom sectors comprise the largest sectors in our sample CLO portfolios. The similarity in terms of sector composition of the largest sectors across CLO portfolios is striking. Industrials ranked first or second among 39 of the 40 portfolios. Similarly, services ranked second, third or fourth in nearly all portfolios and media ranked second, third, or fourth largest sector in all but six portfolios. The maximum median deviation between the composition of a sector in a portfolio and the average composition of that sector across all portfolios was -0.87% and averaged -0.17% across all the sectors. We compared the average sector composition of CLO portfolios in our sample with the composition of new issue leveraged loans during 2004 and 2005 (Exhibit 4). The similarity between the two is striking and yet, given the participation of CLOs in the new issue market described earlier, not entirely shocking. It illustrates that given the limited supply of collateral in the new issue market, portfolio managers ramping up portfolios are constrained to investing in the available assets in the new issue market and cannot afford to be too choosy. Still, the similarity of sector exposures across CLO portfolios is a risk investors should be mindful.
Please see additional important disclosures at the end of this report.
381
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 56
exhibit 4
Comparing Sector Composition: CLO Portfolios and New Issue Leveraged Loans (2004-05)
Industry Distribution: New Issue Leveraged Loans (2004-05)
Utilties 5.8%
Leisure Oil & Gas 3.6% 5.8%
Telecom 7.6%
Building Materials 3.6%
Other 15.3%
Healthcare 7.4%
Media 9.8%
Services/Retail 15.1%
Electronics 6.8% Industrial 19.3%
Industry Distribution: Aggregate CLO Portfolios (2004-05 )
Utilties 4.6%
Leisure 8.5%
Oil & Gas 4.9%
Telecom 5.5%
Building Materials 6.2%
Other 14.1%
Healthcare 7.0% Services/Retail 13.2%
Media 11.3%
Industrial 20.9% Electronics 3.6%
Source: Morgan Stanley, S&P LCD and Intex
Our colleagues, Brian Arsenault and Young Sup-Lee recently explored how economic measures of pricing power and inputs costs are directly linked to margin growth and credit performance and suggests that tougher times are ahead for the high yield market (see “No Margin for Error”, November 10, 2005). They point out autos, consumer, retail, technology, containers and energy as sectors particularly susceptible to margin contraction due to limited pricing power and higher input costs. On the other hand, healthcare, cable/media, metals/mining, chemicals and lodging fell into the category of sectors with pricing power and most insulated from margin contraction. The sector composition of our sample CLO portfolio appears to be evenly split between the vulnerable sectors and those considered to be comparatively resilient. While some of the larger sectors – retail, autos, containers (included in industrials in our classification), oil & gas and electronics – are exposed to margin contraction, CLO investments also have a fair share of relatively stable sectors like media, healthcare, leisure, metals & mining and chemicals (included in industrials).
382
OVERLAPPING OBLIGOR CONCENTRATION
In Exhibit 5, we show the 25 largest obligor concentrations in our sample of 40 CLO portfolios, ranked by the size of the outstanding exposure. 95% of the CLO portfolios in our sample had exposure to Sungard Data Systems, a widely publicized LBO from earlier during the year. Each of the largest 25 obligors appeared in at least half of the portfolios, suggesting a high degree of overlap across portfolios which concerns us. exhibit 5 Issuer Sungard Data Systems
Top 25 Issuers in CLO Portfolios (2004-05) No. of CLOs with Exposure 38
% of Sample 95%
General Growth Properties
34
85%
Kerr-McGee
32
80%
MGM Mirage
29
73%
Panamsat
29
73%
Graham Packaging
24
60%
Charter Communications
21
53%
Allied Waste
28
70%
Warner Chilcott
27
68%
BCP
26
65%
Texas Genco
29
73%
Smurfit-Stone Container
24
60%
Rockwood Specialties Group
25
63%
Movie Gallery
26
65%
Regal Cinemas
28
70%
United Pan-Europe Comm
23
58%
Huntsman
28
70%
Goodyear Tire & Rubber
25
63%
Fidelity National Information
25
63%
Jarden
21
53%
Reliant Energy
22
55%
MCC Iowa
27
68%
R.H. Donnelley
25
63%
Jean Coutu Group
23
58%
Constellation Brands
26
65%
Source: Morgan Stanley and Intex
Please see additional important disclosures at the end of this report.
383
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 56
NO RATINGS BAR-BELLS
The ratings distribution of CLO portfolios in our sample is at Annex 2. Exhibit 6 shows the aggregate ratings distribution of the assets in our sample. Nearly 75% of sample portfolios have assets rated B+ or below, with B+ and B rating categories accounting for about 30% each. Given the typical CLO WARF of 2200-2400, the average quality of the portfolios is not surprising. It is heartening to note that we do not notice significant “bar belling” of portfolios to achieve the required WARF. exhibit 6
BB+ 1.5%
Aggregating Ratings Distribution of CLO Portfolios (2004-05)
B10.2%
BB 6.2%
BB16.8%
CCC & lower 3.2%
Other 1.1%
B 29.3% B+ 31.8%
Source: Morgan Stanley and Intex
The differences in the ratings composition across CLO portfolios are small. The median deviation between the ratings bucket in a portfolio and the average across all portfolios in the sample for that ratings category ranged between -0.8% and 0.2%. Considering that a vast majority of these transactions permitted CCC/Caa buckets ranging from 3.5% to 15%, it is interesting to note the relatively small portion of the sample portfolios were rated CCC or lower.
384
UTILIZATION OF SECOND LIEN, DIP LOAN AND BOND BUCKETS
Over the last 18 months, there has been a growing trend towards CLO portfolio guidelines providing for larger second liens, DIP loans and bond buckets. Second liens and bonds are likely to have lower recoveries than first lien loans due to their higher ranking in the capital structure. We have noted in our previous work (Chapter 49) that the superior performance of CLOs over time has been mainly due to high recovery rates experienced over time. As second lien loans and unsecured bond buckets increase, the risk of CLOs appearing more like the erstwhile CBOs with their mixed track record, begins to rise. On the other hand, there is a potential for attractive spread pick-up with second liens and some unsecured assets. The risk-reward consequences for different tranches can be very different and vary depending on how large the difference is between first lien loan spreads and second lien spreads. Given this background, we examined the utilization of the buckets for second lien loans, bonds and DIP loans in our sample of CLO portfolios (Annex 3). There is a fair degree of variability across portfolios regarding both the provisions for each of these classes of assets and their utilization. Second Lien Loan Buckets: In terms of second liens, 32 of the 40 transactions examined had an explicit second lien bucket ranging from 5% to 10%. In some of the transactions, there were no explicit buckets provided for but the portfolios nevertheless included some allocation to second liens. The average utilization rate of second lien buckets was about 25% and it exceeded 70% only in three of the 40 portfolios examined. DIP Loan Buckets: In terms of DIP loans buckets, the range of provisions is wider ranging from 3% to 15%. However, in terms of their utilization, CLO portfolios appear to be uniformly low, averaging merely 7% with only two portfolios with rates of about 40%. In nearly two thirds of the portfolios, the utilization rate of the DIP loan buckets was zero. Bond Buckets: In terms of bond buckets, the number of CLOs that provide for a bond bucket explicitly is limited. Fifteen of the 40 CLOs in our sample did not have a specified bucket for bonds and one transaction explicitly did not allow for bonds. Where allowed, bond buckets ranged from 5% to 15%, with average utilization rate of about 21%. However, there was a wide dispersion in the utilization rates The limited utilization of second lien and bond buckets indicate that some of the concerns raised in this context might be overblown. We do not suggest that utilization rates will continue to remain as low as they seem to be in our sample but given how tight loan spreads have been recently, now would have been as good a time as any for these buckets to be utilized. That they have not been used as much as they could have been suggests that CLOs have continued to remain CLOs.
Please see additional important disclosures at the end of this report.
385
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 56
CLO EXPOSURE TO RECENT LBO ACTIVITY
As our colleagues, Greg Peters and Rizwan Hussein noted in their recent report (“Those Pesky Barbarians”, November 10, 2005), there is unlikely to be a slowdown in LBO activity any time soon and CLOs are increasingly being seen as important drivers, providing fuel to the momentum in LBO activity. We scoured data on Intex to estimate the exposure of CLOs to some recent LBOs. We counted all direct loan exposure within the broad universe of CLO portfolios to these LBOs (Exhibit 7). Based on this snapshot, it seems that CLO portfolios have been significant resting places for a number of recent LBOs, which raises concerns down the road. exhibit 7
Direct CLO Exposure ($ Million) 579
# of CLOs with Direct Exposure 100
SunGard Data Systems Inc
565
104
PanAmSat Corp
452
93
Texas Genco Holdings Inc
318
117
LBO Metro-Goldwyn-Mayer Inc
LNR Property Corp
315
79
Boise Cascade Corp-Paper Asset
240
133
Loews Cineplex Entertainment
215
88
Nortek Holdings Inc
207
106
US Oncology Inc
178
93
Mid-Western Aircraft Systems
155
99 74
Iasis Healthcare Corp
137
Neiman Marcus Group Inc
127
52
Select Medical Corp
118
49
Source: Morgan Stanley and Intex
386
CLO Exposure to LBOs
CONCLUSIONS
This analysis serves to raise some warning signs as well as alleviate some concerns. The limited variability among CLO portfolios in terms of sector and ratings distributions is certainly a concern. The substantial issuer and sector overlap suggests that in the event of a downturn in credit cycle, diversification across managers may not be sufficient. We have not examined in this chapter if the extent of overlap we find here is also across different vintages, a topic we will address in the near future. Increasingly, LBOs are looking to the leveraged loan markets. The preponderant exposure of CLO portfolios to recent LBOs is a good indication. Further, leverage is rising within LBOs. According to S&P LCD data, the average debt-to EBITDA multiple of large LBOs at the end of the third quarter is 6.1x, up from 5.1x in the first half, 4.9x last year and higher than even the 1997 high of 5.9x. Again, according to S&P LCD data, the average LBO purchase multiple jumped to 8.5x from 7.4x in the six preceding quarters. While, the number of the LBO loans are ending up in CLO portfolios raises a warning flag in our minds, it is important to note that credit curves for LBOs are steep, which suggests that near-term risks are relatively modest. On the other hand, much of the investor rationale for CLO investments remains valid. Near term, technicals are certainly leaning in the direction of CLOs. We expect to see continued robust issuance activity in this very hot corner of corporate credit market. Still, it would pay to evaluate the sector and obligor overlap and be attentive to the LBO activity. We acknowledge the contributions of Simmi Sareen to this chapter.
Please see additional important disclosures at the end of this report.
387
388 12/15/04 7/27/04 8/17/04 3/24/05 6/30/04 12/8/04 12/1/04
Chatham Light CLO
First CLO 2004-1
Galaxy III
Galaxy IV
Grayston II
Green Lane CLO
Hewett's Island CLO II
1/27/05 5/12/04
Centurion CDO 8 Ltd.
Champlain
7/8/04 3/18/04
Carlyle High Yield Partners VI
Celerity
6/24/04
10/20/04
Canyon Capital CLO 2004-1
Blackrock Senior Income Series
22%
3/17/05 5/12/04
Babson CLO 2005-1
Boston Harbor CLO 2004-1
17%
16%
11%
10%
10%
10%
11%
13%
12%
11%
8%
21%
14%
14%
14%
12/20/04
Avenue CLO Fund
12%
2/24/05
Avalon Capital Ltd-3
8%
14%
22%
22%
22%
18%
24%
22%
26%
24%
29%
22%
17%
17%
23%
18%
15%
19%
30%
22%
24%
2%
6%
4%
4%
3%
4%
6%
3%
3%
2%
3%
2%
3%
4%
2%
1%
4%
6%
8%
8%
9%
7%
14%
16%
16%
10%
14%
9%
15%
16%
12%
13%
10%
10%
5%
14%
10%
15%
7%
8%
4%
4%
6%
6%
10%
5%
7%
10%
11%
6%
7%
8%
10%
10%
4%
9%
6%
6%
5%
9%
8%
6%
4%
2%
8%
5%
5%
6%
3%
6%
6%
10%
7%
6%
6%
3%
6%
3%
8%
8%
6%
1%
3%
2%
5%
1%
2%
4%
4%
4%
4%
6%
1%
3%
6%
5%
5%
8%
12%
10%
12%
9%
12%
6%
10%
10%
9%
5%
8%
6%
11%
12%
10%
7%
1%
7%
3%
5%
4%
6%
5%
6%
6%
4%
8%
1%
3%
7%
3%
6%
4%
2%
4%
7%
3%
2%
6%
4%
5%
6%
4%
5%
5%
4%
4%
6%
6%
7%
4%
4%
10%
3%
19% 100%
17% 100%
23% 100%
11% 100%
12% 100%
15% 100%
11% 100%
10% 100%
14% 100%
14% 100%
14% 100%
21% 100%
15% 100%
14% 100%
11% 100%
17% 100%
8% 100%
13% 100%
10% 100%
3/29/05 10/27/04
Ares IX CLO
Atrium III
Closing Services/ Oil and Building Date Retail Industrial Electronics Media Healthcare Telecom Utilities Leisure Gas Materials Other Total 3/23/05 13% 9% 4% 12% 7% 5% 10% 14% 7% 11% 9% 100%
Sector Composition of CLO Portfolios (November 2005)
chapter 56
Name AMMC CLO IV
annex 1
Structured Credit Insights – Instruments, Valuation and Strategies
Please see additional important disclosures at the end of this report.
389
4/21/05
WhiteHorse II CLO
Source: Morgan Stanley, Intex
7/29/04
12/21/04
Whitehorse I
Whitney 2004-1
8/26/04 8/26/04
Stone Tower II
Venture IV
16%
19%
15%
20%
11%
11%
16%
9/22/04 3/16/05
Modena
Stanfield Vantage CLO
11%
5/15/05
Southfork CLO I
16%
10%
10%
5/13/04
3/9/04
10/27/04
14%
15%
5/26/04 5/4/04
8%
13%
4/21/05
6/3/04
11% 11%
Pacifica III
Olympic CLO I
Flatiron CLO 2004-1 Ltd
Northwoods Capital IV
Mountain Capital III
Madison Park Funding I
Long Grove CLO
3/29/05 11/9/04
Landmark CLO V
LCM - II CLO
18%
12%
17%
18%
19%
20%
15%
21%
28%
19%
19%
20%
22%
36%
25%
23%
20%
24%
15%
17%
21%
3%
3%
2%
3%
3%
5%
4%
5%
2%
4%
4%
0%
4%
3%
3%
6%
1%
3%
9%
10%
7%
13%
6%
18%
10%
9%
9%
9%
14%
16%
10%
8%
10%
11%
15%
12%
10%
12%
10%
5%
3%
6%
6%
9%
7%
6%
5%
4%
6%
6%
4%
9%
7%
9%
6%
5%
9%
5%
3%
8%
4%
6%
5%
3%
8%
5%
3%
4%
2%
5%
4%
4%
6%
10%
7%
8%
3%
4%
8%
4%
4%
4%
4%
7%
6%
10%
1%
6%
3%
5%
3%
2%
7%
5%
5%
9%
7%
9%
7%
10%
5%
4%
10%
12%
10%
12%
10%
3%
8%
9%
5%
5%
5%
11%
5%
9%
5%
9%
5%
8%
4%
4%
5%
5%
6%
1%
4%
5%
7%
2%
6%
4%
4%
6%
9%
9%
8%
5%
6%
6%
10%
6%
10%
5%
6%
6%
7%
10%
3%
8%
8%
4%
16% 100%
16% 100%
9% 100%
13% 100%
17% 100%
13% 100%
13% 100%
11% 100%
14% 100%
11% 100%
15% 100%
9% 100%
23% 100%
17% 100%
12% 100%
15% 100%
20% 100%
19% 100%
5% 100%
9/15/04 10/21/04
Katonah VI
Landmark -IV CLO 2004
Closing Services/ Oil and Building Date Retail Industrial Electronics Media Healthcare Telecom Utilities Leisure Gas Materials Other Total 6/29/05 11% 12% 7% 7% 10% 6% 6% 9% 8% 10% 15% 100%
Sector Composition of CLO Portfolios (November 2005)
Name Jasper CLO
annex 1 (cont.)
390 3/29/05
0%
1% 0%
6/30/04 12/8/04 12/1/04
Grayston II
Green Lane CLO
Hewett's Island CLO II
3%
0%
0%
8/17/04 3/24/05
2%
Galaxy IV
7/27/04
0%
2%
2%
0%
1%
2%
2%
2%
3%
2%
0%
2%
BB+ 4%
Galaxy III
12/15/04
First CLO 2004-1
Champlain
Chatham Light CLO
1/27/05 5/12/04
Centurion CDO 8 Ltd.
7/8/04 3/18/04
Carlyle High Yield Partners VI
Celerity
6/24/04
10/20/04
Canyon Capital CLO 2004-1
Blackrock Senior Income Series
3/17/05 5/12/04
Boston Harbor CLO 2004-1
12/20/04
Avenue CLO Fund
Babson CLO 2005-1
2/24/05
10/27/04
Avalon Capital Ltd-3
Atrium III
Closing Date 3/23/05
11%
0%
9%
6%
6%
7%
3%
5%
7%
5%
5%
3%
5%
10%
7%
6%
7%
9%
7%
BB 11%
23%
4%
21%
25%
19%
13%
15%
21%
18%
17%
23%
13%
13%
31%
17%
11%
18%
15%
17%
BB20%
32%
22%
30%
25%
32%
34%
32%
38%
33%
31%
29%
26%
34%
30%
33%
35%
39%
29%
27%
B+ 26%
22%
35%
27%
22%
24%
37%
30%
24%
26%
35%
28%
37%
28%
21%
32%
34%
23%
30%
31%
B 28%
6%
35%
11%
16%
14%
5%
15%
8%
11%
6%
7%
14%
4%
3%
6%
9%
9%
12%
13%
B10%
0%
4%
0%
4%
5%
3%
5%
3%
3%
3%
6%
6%
0%
1%
3%
2%
3%
5%
2%
CCC & lower 2%
Ratings Composition of CLO Portfolios (November 2005)
3%
0%
1%
2%
1%
0%
0%
0%
0%
1%
0%
0%
1%
0%
0%
0%
0%
0%
0%
IG 0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
13%
0%
0%
0%
0%
0%
0%
NR 0%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
Grand Total 100%
chapter 56
Ares IX CLO
Name AMMC CLO IV
annex 2
Structured Credit Insights – Instruments, Valuation and Strategies
Please see additional important disclosures at the end of this report.
391
9/15/04
Stone Tower II
Venture IV
4/21/05
WhiteHorse II CLO
Source: Morgan Stanley, Intex
7/29/04
Whitehorse I
12/21/04
8/26/04 8/26/04
Stanfield Vantage CLO
Whitney 2004-1
9/22/04 3/16/05
Modena
5/13/04 5/15/05
Southfork CLO I
3/9/04
10/27/04
Pacifica III
Olympic CLO I
Flatiron CLO 2004-1 Ltd
5/4/04
5/26/04
Mountain Capital III
Northwoods Capital IV
4/21/05
Madison Park Funding I
6/3/04
LCM - II CLO
Long Grove CLO
3/29/05 11/9/04
Landmark CLO V
10/21/04
Katonah VI
Landmark -IV CLO 2004
Closing Date 6/29/05
Name Jasper CLO
annex 2 (cont.)
0%
1%
2%
2%
3%
2%
0%
1%
4%
3%
2%
0%
1%
0%
1%
2%
0%
4%
2%
BB+ 2%
5%
8%
10%
4%
14%
5%
8%
3%
11%
11%
7%
3%
5%
5%
4%
3%
0%
3%
6%
BB 3%
17%
15%
22%
13%
16%
17%
19%
13%
20%
14%
13%
10%
16%
15%
18%
25%
15%
15%
19%
BB9%
36%
39%
28%
32%
24%
32%
39%
22%
32%
29%
34%
31%
49%
32%
32%
42%
33%
32%
28%
B+ 24%
35%
29%
23%
37%
19%
33%
23%
38%
24%
26%
37%
28%
28%
30%
30%
22%
35%
34%
26%
B 42%
6%
6%
10%
9%
4%
8%
7%
18%
8%
13%
5%
19%
2%
11%
12%
4%
14%
11%
15%
B15%
1%
1%
5%
3%
2%
4%
5%
4%
1%
3%
3%
10%
0%
7%
3%
1%
3%
2%
4%
CCC & lower 5%
Ratings Composition of CLO Portfolios (November 2005)
0%
0%
0%
0%
2%
0%
0%
2%
0%
0%
0%
0%
0%
1%
1%
0%
0%
0%
0%
IG 0%
0%
0%
0%
0%
16%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
NR 0%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
Grand Total 100%
392 6/30/04 12/1/04 6/29/05
Hewett's Island CLO II
Jasper CLO
3/24/05
Galaxy IV 12/8/04
8/17/04
Galaxy III
Green Lane CLO
7/27/04
Grayston II
12/15/04
First CLO 2004-1
5/12/04
Champlain
Chatham Light CLO
1/27/05
Centurion CDO 8 Ltd.
7/8/04 3/18/04
Carlyle High Yield Partners VI
Celerity
6/24/04
10/20/04
Canyon Capital CLO 2004-1
Blackrock Senior Income Series
5/12/04
Boston Harbor CLO 2004-1
5.0%
5.0%
10.0%
10.0%
NA
10.0%
15.0%
10.0%
10.0%
7.5%
5.0%
5.0%
10.0%
15.0%
15.0%
5.0%
10.0%
12/20/04 3/17/05
Avenue CLO Fund
Babson CLO 2005-1
5.0%
10.0%
2/24/05
10/27/04
Avalon Capital Ltd-3
Atrium III
3/29/05
5.0% 10.0%
3/23/05
Maximum Second lien%
Ares IX CLO
Closing Date
1.7%
0.0%
7.4%
1.2%
5.3%
6.1%
0.0%
5.9%
0.0%
0.0%
0.0%
0.0%
0.0%
2.8%
1.1%
2.4%
3.0%
0.0%
3.1%
0.0%
0.8%
Actual Second lien%
7.5%
5.0%
5.0%
5.0%
7.5%
7.5%
7.5%
5.0%
5.0%
7.0%
7.5%
5.0%
7.5%
5.0%
5.0%
5.0%
5.0%
NA
5.0%
5.0%
3.0%
Maximum DIP Loan %
0.0%
0.0%
0.0%
0.0%
1.6%
1.2%
1.1%
0.0%
0.0%
0.0%
1.8%
2.1%
0.0%
0.0%
0.0%
0.7%
0.5%
0.0%
0.0%
0.0%
1.1%
Actual DIP Loan %
Utilization of Second Lien, DIP Loan and Bond Buckets
7.5%
5.0%
5.0%
NA
5.0%
5.0%
NA
10.0%
NA
NA
NA
5.0%
10.0%
10.0%
NA
NA
5.0%
3.0%
10.0%
15.0%
0.0%
Maximum Bond %
0.0%
0.3%
1.7%
0.0%
0.3%
0.0%
0.0%
6.0%
0.8%
2.7%
0.0%
0.9%
2.4%
1.5%
0.0%
0.0%
1.0%
0.0%
5.7%
5.3%
0.0%
Actual Bond %*
chapter 56
AMMC CLO IV
Name
annex 3
Structured Credit Insights – Instruments, Valuation and Strategies
Please see additional important disclosures at the end of this report.
393
8/26/04
WhiteHorse II CLO
NA
NA
10.0%
10.0%
5.0%
5.0%
5.0%
5.0%
5.0%
NA
10.0%
NA
NA
10.0%
5.0%
5.0%
NA
NA
NA
Maximum Second lien%
1.3%
2.1%
2.9%
0.0%
1.1%
2.6%
4.1%
4.3%
0.0%
0.0%
4.1%
0.0%
1.5%
3.9%
0.0%
0.0%
0.0%
0.0%
3.5%
Actual Second lien%
5.0%
5.0%
5.0%
7.5%
5.0%
3.0%
3.0%
5.0%
5.0%
5.0%
3.0%
15.0%
5.0%
5.0%
7.5%
5.0%
5.0%
5.0%
5.0%
Maximum DIP Loan %
*"Actual Bond%" does not include senior secured notes which in some transactions may be classified as bonds. **NA signifies cases where there is no explicit provision for second lien or DIP loans or bonds. Source: Morgan Stanley, Intex
7/29/04 4/21/05
Whitehorse I
12/21/04
Venture IV
Whitney 2004-1
3/16/05 8/26/04
Stone Tower II
Modena
Stanfield Vantage CLO
5/15/05 9/22/04
Southfork CLO I
5/13/04
3/9/04
10/27/04
Pacifica III
Olympic CLO I
Flatiron CLO 2004-1 Ltd
5/4/04
Mountain Capital III
Northwoods Capital IV
4/21/05 5/26/04
Madison Park Funding I
6/3/04
11/9/04
LCM - II CLO
Long Grove CLO
3/29/05
10/21/04
9/15/04
Closing Date
1.0%
0.0%
0.5%
0.0%
0.0%
0.0%
0.0%
0.1%
0.0%
1.5%
0.3%
0.0%
0.0%
0.8%
0.0%
0.0%
0.0%
0.0%
0.0%
Actual DIP Loan %
Utilization of Second Lien, DIP Loan and Bond Buckets
Landmark CLO V
Landmark -IV CLO 2004
Katonah VI
Name
annex 3 (cont.)
5.0%
5.0%
15.0%
NA
5.0%
5.0%
5.0%
7.5%
NA
NA
NA
NA
NA
10.0%
5.0%
NA
5.0%
5.0%
NA
Maximum Bond %
0.0%
0.0%
6.2%
0.7%
0.0%
1.2%
1.3%
1.2%
2.6%
4.0%
1.4%
0.0%
0.0%
3.2%
0.2%
0.0%
1.1%
3.0%
7.8%
Actual Bond %*
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 57
European CLOs: Touring the Continent
January 24, 2006
Primary Analyst: Vishwanath Tirupattur Primary Analyst: Sivan Mahadevan In the past few years, the European loan market has been transformed – from within and from without – in many important ways. The landscape of investors in the primary market for European loans is clearly changing (Exhibit 1). Institutional investors’ share of the primary market has increased about 10 times over the past six years – from 4.2% in 1999 to 40.3% in 2005 (according to S&P). CLOs are the most dominant of the institutional investors, currently accounting for over 70% of the institutional investors in the market. At the same time, the share of European banks has declined from over two-thirds in 1999 to less than half (46.7%) as they have begun to move away from their roles as both originators and investors of corporate loans. exhibit 1
Primary Market for European Loans – Who Are the Players?
European Banks
Non-European Banks
Other Institutional Investors
Securities Firms
CLOs
80% 70% 60% 50% 40% 30% 20% 10% 0% 1999
2000
2001
2002
2003
2004
2005
Source: S&P and Morgan Stanley
Basel II has important implications for bank holdings of leveraged loans and may help to explain the changing role of banks in this space. In general, the new risk-weights work against leveraged loans, and this makes holding them on bank balance sheets far less attractive. On the other hand, exposure to leveraged loans in securitized form such as CLOs is significantly favorable, as is hedging of the loan exposures through instruments such as loan CDS (see the work of our European colleagues Jackie Ineke and Christine Miyagishima, “Leveraged Loans: Suffering Under Basel II,” May 9, 2005). We have also discussed our thoughts on the significance of such secular shifts on the underlying markets (see “2006 Outlook: The Cyclical versus Secular,” December 16, 2005). The European loan market is thriving in terms of record issuance volume and everincreasing transaction size (€123.4.4 billion from 214 deals in 2005 compared to €64.8 billion from 140 deals in 2004). Spreads across the board remain tight. On the other
394
hand, structures have also become increasingly aggressive, with leverage multiples at an all-time high and deteriorating credit quality. As of the end of November 2005, 72.1% of the deals in S&P’s European Leveraged Loan Index were rated B and 22.8% rated BB compared with 46.7% of loans rated BB and 50.1% rated B as of November 2004. LBOs – in particular, secondary buyouts and public-to-private transactions – are the most dominant segment of the European loan market, accounting for €103.5 billion from 192 transactions in 2005, more than double the €43.8 billion mark set in 2004. Given the growing significance of European CLOs, it is useful to analyze the composition of CLO portfolios against the backdrop of changing investor landscape, exuberant loan markets and some signs of gathering credit risks to analyze where CLO managers are placing their bets in terms of sector composition, geographic diversification, credit quality and LBO exposure. Further, it is important to understand the extent of overlap in terms of sectors, countries and obligors from the perspective of investors holding portfolios of CLO tranches. In addition, given the size of mezzanine and bond buckets in CLO portfolios, it is important to understand the extent of their utilization. EUROPEAN CLO ISSUANCE IS BLOOMING
European CLO issuance has been impressive (Exhibit 2). Between 2004 and 2005, issuance more than doubled from €7.96 billion to €16.26 billion. In itself, 2004 issuance was more than two times the 2003 issuance of €3.73 billion. By our count, there were 24 transactions in 2005 and 64 outstanding CLOs in all managed by 33 different managers. The issuance boom in European CLOs is consistent with explosion of investor interest globally in leveraged loans as an asset class and CLOs as an attractive way to get exposure to leveraged loans. Robust primary issuance in the underlying loans, steady performance, growing liquidity, stable ratings and secular changes in terms of Basel II have contributed to the surging popularity of CLOs, including European CLOs. Over past 12 months, there have been no downgrades of European CLO tranches by any of the three rating agencies, and with the exception of two upgrades, ratings have been unchanged for nearly all European CLO tranches, which demonstrates their rating stability. exhibit 2
Growing European CLO Issuance
$ Billions 18 16 14 12 10 8 6 4 2 0 2000
2001
2002
2003
2004
2005
Source: S&P and Morgan Stanley
Please see additional important disclosures at the end of this report.
395
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 57
New issue spreads remain at or close to historical tights. Over the past one year, European CLO funding costs have tightened by about 7 bps and remain tighter by a couple of bps to their US counterparts. As we have noted elsewhere, the ability to lock in non-recourse term funding at such tight levels with the prospect of reinvesting at potentially wider spreads makes the case for CLO equity fairly attractive. A comparison of the average new issuance spreads between US and European leveraged loans suggests that, while on average European loan spreads are wider than their US counterparts, their tenor is longer and leverage levels are higher. DATA
Using Intex data, we examined 26 European CLO portfolios (as of December 2005) of transactions issued during 2003–05, which covers about 60% of the issuance during that period. The sample of deals covered 18 different managers (representing more than half of all European CLO managers) with no manager represented more than twice in the sample data. Given the limited overlap of managers and the diversity of managers and transactions within the sample, we consider that our sample provides a reasonable reflection of the portfolio composition of European CLOs. LOWER DIVERSITY WITHIN BUT HIGHER DISPERSION ACROSS CLO PORTFOLIOS
The collateral composition by sector of the 26 portfolios in our sample is summarized in Annex 1. Regular readers of our research may recall that we recently published a similar analysis of US CLO portfolios (see Chapter 56), in which we found a striking similarity of sector composition across different CLO portfolios as well as with the sector composition of the new issue leveraged loans. In comparison to the US CLO portfolios, we find that the consensus of sector preferences across European CLO portfolios is lower. The services sector ranked first in most number of deals – but only in eight of the 26 portfolios. In six deals, it actually ranked among the smallest two sectors. The retail and cable sectors ranked first or second in the most number of deals. Again, they also ranked among the smallest sectors in a handful of deals. In contrast, the largest sector among the US CLO portfolios – industrials – ranked first or second in nearly all of the 40 portfolios we analyzed. An alternative measure of dispersion entails examining the coefficient of variation (CV) of a sector weight across CLO portfolios. The average CV across all sectors was 56% for European CLOs compared to 40% for the US CLOs we analyzed. This also suggests a greater variability of sector exposures across European CLO portfolios relative to US CLO portfolios. It is important not to confuse a CLO’s sector diversification with the diversification of sectors across CLOs. Diversity scores – the common measure of diversification within a CLO portfolio – are generally lower in European CLOs relative to US CLOs. Our point here is that the sector diversification across CLO portfolios based on our analysis is significantly higher in European CLOs. This is of material significance to investors holding portfolios of CLO tranches. We can think of a few plausible explanations for this notable difference between US and European CLO portfolios. First, given the geographic diversity and the potential for the prevalence of different business conditions and macro economic forces in
396
different countries, the absence of consensus in terms of which sector offers the best value is not entirely surprising. Second, despite their recent decline, banks are still dominant in the European loan market. Not all managers may have equal access to allocations of new issues. Finally, secondary trading is limited in Europe and lacks the depth of its US counterpart. WHERE ARE THE SECTOR BETS?
We also compared the aggregate average sector composition of our sample portfolios with the sector composition of new issue European leveraged loans. We used sector composition of new issue loans reported by S&P for 2004 and the first three quarters of 2005, which broadly represents the universe of new issue loans available for the CLOs in our samples to invest (Exhibit 3). Relative to the new issue loans market, the aggregate CLO portfolio is overweight in the services, manufacturing, printing, and publishing sectors; underweight in telecommunications, computers & electronics, chemicals and food & beverage sectors; and equal-weight in the cable sector. exhibit 3
Comparing Sector Composition: CLO Portfolios and New Issue Leveraged Loans (2004-05)
Sector Composition: Aggregate CLO Portfolios Computers & Electronics 2%
Building Materials 4%
Other 29%
Printing & Publishing 7% Manufacturing 9%
Chemicals 7%
Cable 9% Telecom 6%
Services 13% Retail 10%
Food&Beverage 4%
Sector Composition: New Issue Leveraged Loans Printing & Publishing 6%
Building Materials 5%
Computers & Electronics 4%
Other 23%
Manufacturing 6% Chemicals 8%
Cable 9%
Services 9% Retail 11%
Telecom 14%
Food&Beverage 5%
Source: Morgan Stanley and Intex
Please see additional important disclosures at the end of this report.
397
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 57
Morgan Stanley European credit strategist Neil McLeish favors sectors that enjoy pricing power and is bearish on those with limited pricing power. Among the sectors with significant exposure, in our CLO portfolios, cable, packaging and utilities fall in to the former category of sectors with pricing power. Manufacturing, retail, telecommunications, food & beverage and chemicals fall into the latter as sectors with limited pricing power with a potential for margin compression. As noted above, there is significant variation among CLO portfolios in terms of exposure to sectors with and without pricing power. On an aggregate basis, the sector composition of our sample portfolios appears to be evenly split between sectors vulnerable and those considered comparatively resilient WHERE ARE THE COUNTRY BETS?
The geographic diversity of European CLO portfolios subject to country specific risks and macroeconomic conditions adds another dimension to risk diversification relative to US CLOs. We summarize the country diversification of the CLO portfolios in our sample in Annex 2. exhibit 4
Comparing Country Composition: CLO Portfolios and New Issue Leveraged Loans (2004-05)
Country Composition: Aggregate CLO Portfolios Ireland 0%
Netherlands 4% Spain 3%
North America 6%
Italy 4%
France 17% Other 12%
Germany 32%
UK 22%
Country Composition: New Issue Leveraged Loans
Spain 8%
Netherlands 7%
Ireland 2% Italy 10%
North America 3%
Other 8%
France 16%
Germany 19%
Source: Morgan Stanley and Intex
398
UK 27%
The variation in country exposures across CLO portfolios was not as broad as the variation in the sector distribution across portfolios and generally consistent with the relative sizes of different economies. Germany was the largest country exposure in our sample, ranking within the top three countries in nearly all of the portfolios. The United Kingdom and France ranked within the top three in a large proportion of the portfolios. It is noteworthy that in four portfolios, North America (US and Canada combined) was the largest regional exposure, accounting for over 40% of the portfolio in three cases. Again, we compare the country diversification of the aggregate CLO portfolios from our sample to the country diversification of the European new issue leveraged loans as a proxy for the investable universe for European CLO managers. At an aggregate level, Germany is the largest overweight relative to the new issue market, with the difference being made up by a relative underweight in the UK, Spain, the Netherlands and Italy. OVERLAPPING OBLIGOR CONCENTRATION
In Exhibit 5, we show the 25 largest obligor concentrations in our sample of 26 CLO portfolios, ranked by the size of the outstanding exposure. Springer Science & Business, a German publishing company, and TDF, a French broadcasting company, top the list of obligors – 73% and 81% of the CLOs in the sample had exposure to these two obligors, respectively. While there is a significant amount of overlap (each of the largest 25 obligors appeared in about 40% of the sample deals), in comparison to our US CLO sample, the obligor overlap is somewhat lower, consistent with our finding regarding sector distribution across portfolios. exhibit 5 Obligor Springer Science & Business TDF Aster 1 Kabel BW WAM Acquisition Cognis Deutschland II Kappa Beheer BV Satbirds Finance SARL Frans Bonhomme Grohe GmBH A.T.U Handels GMBH & Saga Tank & Rast Elis Financiere Gaillon Demag Debitel Nachtwache Acq GMBH Materis SEAT Pagine Invensys Dragoco Gerberding World Directories Acquisition Sanitec Europe OY Picard Surgeles
Top 25 Obligors in CLO Portfolios No. of CLOs with Exposure 19 21 15 13 14 20 18 14 18 16 14 12 16 14 12 14 15 19 12 10 13 11 13 13 12
% of Sample 73% 81% 58% 50% 54% 77% 69% 54% 69% 62% 54% 46% 62% 54% 46% 54% 58% 73% 46% 38% 50% 42% 50% 50% 46%
Source: Morgan Stanley and Intex
Please see additional important disclosures at the end of this report.
399
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 57
RATINGS DISTRIBUTION
A unique feature of the European loan market is that a large proportion of the loans are not publicly rated. Rating agencies use “shadow ratings” for their analysis, which are generally estimates provided based on abbreviated analyses. S&P suggests that these shadow ratings may be up to two full notches away from the results of their full public rating analysis. 1 In our sample, about 62% of the assets in the portfolio fall in the unrated category. As noted earlier, there is a clear downward shift in the quality of the European leveraged loans, with the proportion of loans rated B increasing by about 43% between 2004 and 2005. This decline in credit quality combined with the uncertainty of the “true” credit quality implied by the shadow rating is disconcerting to us. It has historically been the case that loan default experience in Europe has been lower than the same in US because of the reluctance of bank lenders to be subjected to the interference of courts. Bank lenders have traditionally resorted to covenant amendment and refinancing prior to an event of default on a pro-active basis. In fact, in this context the declining significance of banks as investors might be risk additive. We share the concern expressed in the S&P report, 2 that the growing institutionalization of the market with larger syndicates and participations could impede the renegotiation of loan terms. As such, default experience may begin to resemble that of US loans. Thankfully, despite a recent pickup in defaults, our baseline expectations point to a relatively benign default environment in the near term. As we have noted elsewhere, the key to continued strong performance of CLOs – European or US – will be high recoveries. EXPOSURE TO LBOS
The LBO related loans have dominated deal flow in the European loan market, and LBO risk is seen as the key risk to European credit investors. We looked at recently completed as well as pending LBO transactions and exposure of our sample CLO portfolios to them. About 26% of the assets in our portfolios were related to recent LBOs and about 6% to pending LBOs. We show the largest exposure to completed and pending LBOs in our sample portfolios in Exhibit 6. While LBO exposure to CLO portfolios bears watching, it is important to point out that the risk implications are different across different CLO tranches. Senior tranche investors have significant subordination cushion, benefit from structural protections, and may have less to worry about if LBO exposure translates in to a rise in idiosyncratic risks. While risks are higher for lower tranches, the emergence of loan CDS contracts may be useful in hedging some of these risks. In our mind, it is critical to understand the extent of obligor overlap, especially LBO names across CLO tranches, and consider hedging them.
1 2
400
European CDO of Leveraged Loans Q4 2005 Review”, Standard and Poor’s, 12 January 2006. Ibid
exhibit 6
LBO Name France Telecom Cable SA
Top 25 Obligors in CLO Portfolios CLO Exposure to LBO (€ Million) Completed LBOs 145.88
Yellow Brick Road
% of sample with Exposure 81%
67.28
73%
KappAhl AB
111.99
69%
Friedrich Grohe AG & Co KG
101.18
62%
96.35
62%
Autobahn Tank & Rast Holding Debitel AG
82.00
58%
Wam Acquisition SA
116.54
54%
ATU Auto-Teile-Unger GmbH
100.81
54%
73.60
50%
VNU World Directories Groupe OGF
46.40
50%
Saga Group Ltd
98.37
46%
Picard Surgeles SA
71.22
46%
Thule Holding AB
48.64
38%
Viterra AG
47.93
35%
Koninklijke Vendex KBB NV
62.79
27%
Coral Eurobet PLC
43.31
27%
Cory Environmental Ltd
40.81
27%
IMO Car Wash Group Ltd
44.06
23%
VNU World Directories
41.75
23%
Ahold-Supersol Supermarkets
34.56
23%
Baxi Group Ltd
32.54
23%
Terreal SA
30.15
23%
Safety-Kleen Europe Ltd
35.34
19%
Mivisa Envases SA
29.57
19%
TM Group Holdings PLC
25.76
19%
Pending LBOs Frans Bonhomme SA
102.55
69%
Aster City Cable Sp Zoo
120.35
58%
Kabel Deutschland GmbH & Co KG
118.42
50%
Antimex Nordic
67.82
38%
Pets At Home Ltd
18.68
19%
Kwik-Fit Holdings PLC
13.34
12%
Hirschmann Electronic GmbH
14.64
12%
Strix Group
7.54
4%
Spirit Group Ltd.
8.62
4%
FL Selenia SpA
9.35
4%
Source: Morgan Stanley and Intex
Please see additional important disclosures at the end of this report.
401
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 57
UTILIZATION OF MEZZANINE LOAN AND BOND BUCKETS
An important feature of European CLOs is the size of the mezzanine loan buckets. Mezzanine debt in Europe is meant to fill the gap between senior debt and equity within a firm’s capital structure. It ranks behind senior debt in terms of priority but ahead of equity. Uniquely, mezzanine loans benefit from additional upside return (rolled-up interest, warrants, options, etc.) while sharing the underlying secured collateral with the senior debt investors, albeit with a lower priority. The lower claim on the collateral implies likely lower recoveries. The risk-reward consequences of larger mezzanine loan buckets can be different for different CLOs, not unlike the effect of second-lien loans we have discussed in our previous work (see Chapter 50). To some extent, the effects are similar for bond buckets in CLO portfolios. In this light, we look at the sizes of mezzanine loan and bond buckets in European CLOs and examine their utilization rates (Annex 3). Mezzanine Loan Buckets: All of the CLOs in our sample provided for mezzanine loans. The maximum allowable size of mezzanine loan buckets in our sample ranges between 10% to 25%, averaging around 19%. The average utilization rate is 56% across the different portfolios. By way of comparison, the average utilization rate for second lien loans in US CLOs was 25%. Bond Buckets: Over 80% of the CLOs in our sample had explicit bond buckets. The size of unsecured bond buckets ranged between 5% and 20%. The utilization rates vary widely – from 100% to 0%. The fairly high utilization rates of mezzanine loan buckets bodes well for CLO equity tranches and is somewhat negative for the senior debt tranches. CONCLUSIONS
European CLOs are becoming larger players in the European leveraged loan markets. This analysis serves to highlight several of their important attributes in terms of sector, country composition, obligor overlap and LBO risk exposure. We are encouraged that the extent of overlap across CLO portfolios in terms of sectors seems lower in European CLOs, which suggests that diversification across managers is beneficial if there is a downturn in the credit cycle. Similarly, country diversification in European CLO portfolios also provides a unique diversification play. Given the current tight liability spread environment and the ability to lock up non-recourse term funding that CLOs provide, the case for European CLOs appears attractive. This is not to suggest that investor should be complacent about obligor overlap and exposure LBOs. It is important to be cognizant of these risks. The emerging loan CDS market offers a plausible hedging strategy to manage such risks. Developments in that market are worth watching for investors holding portfolios of CLO tranches. The weakening credit quality is yet another reason to consider hedging strategies. Finally, the secular changes being unleashed by Basel II are significant and, in our judgment, bode well for obtaining exposure to European loans in the form of CLO tranches. We acknowledge the contributions of Simmi Sareen to this chapter.
402
Please see additional important disclosures at the end of this report.
403
21-Oct-04 10-Dec-04 16-Dec-03 30-Nov-04 25-Feb-04 24-Apr-03 24-Apr-03 29-Jul-04 19-May-05 11-Sep-03 22-Nov-04 28-Aug-03 10-Mar-05 22-Jul-03 29-Jun-05 15-Apr-04 20-Jan-04 05-Aug-04 07-Apr-04 06-Apr-05 03-Jun-03 14-Sep-04 06-Oct-03 30-Jun-04 24-Jun-04 28-Jul-04
Closing Date
Source: Morgan Stanley, Intex
Adagio Alzette European CLO Avoca CLO 1 Avoca CLO II Clarenville CDO Copernicus Euro CDO 2 BV Duchess II CDO Duchess III CDO Duchess IV CDO Eurocredit CDO 3 Eurocredit CDO IV Galway Bay BV Global Enhanced Loan Fund GSC European CDO I GSC European CDO II Harvest CLO Jubilee III Jubilee IV Leopard II CLO Leopard III CLO North Westerly CLO I North Westerly CLO II Partholon CLO 1 Petrusse CLO RMF Euro CDO II Vallauris CLO
Deal Name
7% 4% 8% 9% 4% 4% 11% 13% 12% 7% 6% 9% 4% 11% 15% 7% 9% 13% 8% 11% 8% 8% 3% 5% 5% 3%
3% 1% 2% 2% 1% 3% 8% 5% 6% 2% 2% 1% 1% 1% 2% 1% 6% 6% 0% 0% 3% 1% 1% 0% 0% 2%
2% 8% 5% 6% 2% 4% 2% 1% 2% 10% 6% 5% 4% 0% 0% 7% 5% 1% 5% 3% 8% 8% 8% 4% 5% 4%
Banking & Building Autos Insurance Materials
annex 1
6% 16% 6% 10% 11% 13% 6% 3% 11% 9% 15% 11% 7% 14% 15% 7% 11% 13% 13% 12% 4% 5% 6% 14% 9% 7%
8% 13% 6% 7% 10% 6% 2% 6% 4% 13% 8% 4% 6% 5% 4% 8% 11% 8% 5% 10% 6% 8% 2% 12% 21% 4%
Cable Chemicals
4% 2% 1% 2% 5% 2% 0% 0% 0% 1% 1% 3% 9% 4% 5% 4% 0% 0% 4% 1% 3% 2% 5% 2% 3% 3%
8% 2% 7% 6% 2% 3% 7% 7% 1% 6% 4% 4% 0% 4% 5% 6% 1% 5% 7% 5% 7% 4% 7% 6% 0% 4%
3% 8% 11% 1% 14% 8% 15% 13% 7% 8% 7% 17% 10% 16% 11% 5% 7% 7% 9% 4% 8% 5% 9% 8% 9% 3%
7% 4% 6% 11% 12% 3% 6% 5% 8% 5% 6% 5% 6% 7% 2% 3% 8% 3% 7% 0% 11% 13% 8% 5% 4% 5%
9% 5% 11% 8% 6% 6% 4% 9% 8% 6% 9% 11% 6% 4% 7% 8% 10% 11% 8% 12% 4% 8% 7% 5% 1% 7%
12% 7% 17% 16% 0% 11% 7% 10% 9% 8% 10% 11% 4% 2% 5% 15% 14% 10% 8% 11% 17% 16% 13% 6% 16% 13%
14% 12% 13% 14% 6% 17% 13% 6% 11% 15% 16% 13% 10% 10% 12% 17% 3% 5% 9% 14% 7% 8% 19% 10% 10% 28%
7% 3% 1% 2% 12% 7% 1% 8% 7% 1% 1% 4% 20% 12% 11% 6% 8% 8% 10% 11% 1% 1% 3% 5% 3% 2%
4% 3% 0% 0% 4% 1% 3% 4% 3% 2% 4% 1% 6% 0% 0% 0% 3% 3% 1% 2% 4% 3% 0% 3% 1% 4%
Computers & Food & Printing & Electronics Beverage Manufacturing Packaging Publishing Retail Services Telecom Utilities
Sector Composition of CLO Portfolios
6% 12% 7% 5% 11% 13% 15% 11% 11% 8% 5% 2% 7% 11% 8% 7% 5% 6% 6% 6% 9% 11% 10% 15% 13% 13%
Other
Total
100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100%
France 12% 10% 20% 11% 13% 8% 27% 23% 11% 10% 9% 10% 17% 11% 10% 17% 17% 7% 9% 23% 10% 17% 42% 0% 0% 200%
404 Germany 20% 19% 21% 15% 22% 24% 18% 23% 29% 17% 33% 34% 25% 31% 30% 29% 31% 32% 37% 24% 21% 28% 21% 0% 0% 1300%
Ireland 0% 5% 1% 1% 0% 0% 2% 1% 10% 1% 4% 0% 0% 1% 0% 1% 0% 2% 3% 3% 1% 1% 1% 0% 0% 0%
Italy 0% 2% 2% 3% 1% 3% 4% 2% 8% 4% 7% 5% 7% 3% 5% 4% 4% 0% 0% 2% 1% 5% 3% 0% 0% 0%
Netherlands 5% 4% 17% 8% 14% 12% 12% 9% 5% 6% 11% 11% 8% 10% 8% 3% 4% 28% 19% 18% 4% 7% 5% 0% 0% 0%
Other 6% 8% 11% 8% 4% 9% 6% 10% 9% 6% 11% 16% 11% 11% 9% 19% 12% 13% 9% 8% 5% 7% 6% 0% 0% 0%
Sector Composition of CLO Portfolios Spain 2% 2% 0% 3% 7% 6% 2% 2% 7% 3% 0% 0% 7% 4% 1% 5% 3% 3% 3% 7% 2% 2% 6% 0% 0% 0%
UK 11% 22% 12% 50% 38% 37% 22% 24% 18% 7% 21% 21% 24% 20% 28% 19% 22% 15% 17% 13% 6% 17% 10% 0% 0% 400%
North America 44% 29% 16% 0% 1% 1% 6% 5% 2% 47% 4% 5% 1% 6% 9% 4% 6% 0% 3% 3% 50% 16% 7% 0% 0% 400%
Grand Total 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 0% 0% 0%
chapter 57
Source: Morgan Stanley, Intex
Deal Name Adagio Alzette European CLO Avoca CLO 1 Avoca CLO II Clarenville CDO Copernicus Euro CDO 2 BV Duchess II CDO Duchess III CDO Duchess IV CDO Eurocredit CDO 3 Eurocredit CDO IV Galway Bay BV Global Enhanced Loan Fund GSC European CDO I GSC European CDO II Harvest CLO Jubilee III Jubilee IV Leopard II CLO Leopard III CLO North Westerly CLO I North Westerly CLO II Partholon CLO 1 Petrusse CLO RMF Euro CDO II Vallauris CLO
annex 2
Structured Credit Insights – Instruments, Valuation and Strategies
annex 3
Deal Name Adagio
Utilization of Mezzanine Loan and Bond Buckets
Maximum Mezz Loan % 15%
Actual Mezz. Loan % 14%
Maximum Bond % 10%
Actual Bond % 0%
Alzette European CLO
10%
3%
5%
0%
Avoca CLO 1
20%
16%
NA
0%
Avoca CLO II
20%
11%
NA
0%
Clarenville CDO
10%
2%
10%
9%
Copernicus Euro CDO 2 BV
20%
9%
NA
6%
Duchess II CDO
25%
18%
0%
0% 0%
Duchess III CDO
25%
16%
0%
Duchess IV CDO
20%
12%
2%
0%
Eurocredit CDO 3
15%
12%
5%
0%
Eurocredit CDO IV
20%
8%
10%
6%
Galway Bay BV
25%
14%
10%
0%
Global Enhanced Loan Fund
10%
2%
5%
5%
GSC European CDO I
20%
10%
5%
0%
GSC European CDO II
20%
8%
5%
0%
Harvest CLO
20%
16%
5%
0%
Jubilee III
25%
19%
3%
0%
Jubilee IV
25%
14%
5%
0%
Leopard II CLO
20%
16%
8%
2%
Leopard III CLO
25%
13%
5%
2%
North Westerly CLO I
20%
9%
20%
0%
North Westerly CLO II
15%
7%
5%
0%
Partholon CLO 1
10%
9%
NA
0%
Petrusse CLO
10%
3%
NA
0%
RMF Euro CDO II
20%
10%
20%
10%
Vallauris CLO
25%
8%
5%
0%
Source: Morgan Stanley and Intex
Please see additional important disclosures at the end of this report.
405
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 58
Dana + Delphi: Watch the Count and the Roll
March 10, 2006
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Vishwanath Tirupattur Primary Analyst: Pinar Onur While the Delphi default last fall was worthy of an overpowering “strike one” to the billions of dollars of junior mezzanine risk out there, a higher than average recovery weakened the pitch somewhat. Dana's bankruptcy filing last week was perhaps a weaker strike, given that structured credit exposure is a fraction of Delphi’s. Furthermore, recovery could be in the same range, if not higher (bonds are trading north of $70). In a game where it takes three strikes for bad things to happen, Delphi plus Dana amounts to less than one strike even to the most exposed junior mezzanine investors. Nevertheless, some more recent structures could be behind on the count now as they have not had enough time to build it up in their favor (i.e., the passage of time = balls). However, we note that the more recent deals are less likely to contain one or both credits. While the Dana bankruptcy filing has been received quite well in the credit markets, we are concerned about potential further events that could raise the risk of tranche downgrades. In particular, we believe credits to watch (that are in synthetic CDOs) include Lear (rated BBB- for a solid two years until last summer), American Axle (currently BBB- and has been IG for over two years), and, of course, GM Co., which we believe is proportionately over-represented in bespokes relative to its much bigger subsidiary, GMAC. We noted that Delphi was not enough to slow down the structured credit bid (see Chapter 53), and adding Dana is not making much of a difference either (our early read), but further events could change the chemistry, especially for longer-dated transactions. Six months ago, it was difficult to envision a single event derailing the structured credit market. Today, the recent spate of defaults could be priming the market for a sentimentturning event. Credit markets have become addicted to the structured credit bid and even a modest slowing of it could impact underlying spreads significantly.
406
In the more liquid corner of the structured credit market, all the talk is equity tranches, with implied correlations now as low as they were during last spring’s correlation event (although equity prices are still 15 to 20 points higher than they were then). Even though none of the defaulted credits have been in the on-the-run IG indices (CDX 4 or 5), clearly the fear is helping to re-price equity and even some mezzanine risk, at least on a correlation basis. Yet, we remind investors that much of the support brought to the equity tranches last spring was from absolute takers of risk, who are looking more at upfront prices than correlation. It is important to note that mezzanine flows both in indices and bespokes continue to be very strong, as structured credit appetite continues unabated for now. In short, the cyclical Dana event is being overshadowed by the secular shifts (see our “2006 Outlook: The Cyclical vs. the Secular,” December 16, 2005). And on a related note, the Dana bankruptcy makes the upcoming roll on the HY indices even more interesting from a tranche perspective (see last section of this chapter). exhibit 1
Dana vs. Delphi: CDO Exposure Delphi 842
Dana 315
% of S&P Universe
35%
13%
S&P Actions or Estimates
278
32
S&P Rated IG/HY CBOs
15
61
S&P Rated CLOs
51
16
S&P Rated Synthetic CDOs w/Exposure
Moody’s Rated Synthetic CDOs w/Exposure Moody’s rated IG/HY CBOs Moody’s rated CLOs
230
24
70
79
Inc above
17
Source: Morgan Stanley, S&P, Moody’s.
IMPACT ON SYNTHETIC CDOS
Notwithstanding that 13% of all S&P rated synthetic CDOs have exposure to Dana, S&P considers that the ratings migration of Dana has already been reflected in existing ratings on these CDOs, and as such, any incremental impact is likely limited. About a quarter of these transactions may be placed on negative watch and possibly downgraded. We caution though that the rated portion of the synthetic CDO market is only a fraction of all synthetic CDOs. As a comparison, Delphi was a much bigger event, both in terms of CDO exposure and ratings actions (see Exhibit 1).
Please see additional important disclosures at the end of this report.
407
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 58
LIMITED IMPACT ON CASH CDOS
According to S&P, Dana exposure in cash CDO portfolios amounts to approximately $211 million. In over 90% of cash CDOs with Dana, the exposure amounts to less than 3% of their portfolios. The Dana bankruptcy is a non-event for CLOs both because it appears in relatively few CLO portfolios and, post-default, the outstanding loan has traded above par. Legacy HY CBOs and, to a lesser extent, IG CBOs have some Dana exposure but according to S&P, less than 10% of these transactions may experience a negative rating action as a consequence of the Dana bankruptcy filing. Interestingly, some ABS CDOs have exposure to Dana, albeit indirect, through synthetic CDO tranches in their portfolios. Still, thanks to the second order nature of such exposure, we do not expect Danacaused negative ratings actions in ABS CDOs. exhibit 2
Dana Ratings and Spreads
Rating
5yr CDS (bp) 0
BBB/Baa2 12 BBB-/Baa3 11
500
BB+/Ba1 10 BB/Ba2 9
1,000
BB-/Ba3 8 B+/B1 7 B/B2 6
1,500
B-/B3 5
2,000
CCC+/Caa1 4
S&P
CCC/Caa2 3
Moody's 2,500 3,000 450
CCC-/Caa3 2
5yr CDS (bp)
CC/Ca 1 Default 0
400
350
300
250 200 150 Days from Filing
100
50
0
Source: Morgan Stanley
DANA AND THE TALE OF BBB- CREDITS
While Dana has largely been a high yield credit since 2001 (see Exhibit 2), it did trade as tight as 100 bp for a period of time and was rated BBB- by S&P for nine months until September of last year, qualifying it as a fallen angel default, at least by S&P’s rulebook. Dana found itself into many S&P-rated CDOs, and in many ways it bolsters S&P’s point about BBB- credits being much more risky than BBBs (it defaulted less than six months after being rated BBB-). We highlight that S&P CDO model changes were implemented to reflect exactly this phenomenon (see Chapter 67). While underlying credit ratings should not be the sole driver of credit selection, there is an important lesson to be learned here.
408
WATCH FOR OTHER CREDITS AND A JUMP IN THE COUNT
As with the airline defaults last year, the underlying structural issues for the auto industry have the potential to create a powerful link among the remaining credits in the sector (or it could serve to bifurcate them). Any such link would only be exacerbated by the pension issues faced by these entities. As we highlighted earlier (see Chapter 51) of the 28 names with unfunded pension obligations amounting to a significant proportion to their market capitalization in early 2005, 13 were auto or auto part manufacturers. A GM event (without GMAC) could quickly move the count even further behind for some of the newer structures. Even an event involving other entities present in legacy investment grade deals or high yield entities could have an impact on structured credit investor sentiment to the extent these events generate concern over a domino effect that results in a restructuring of the entire US auto industry (à la legacy airlines). That is the key risk that could materialize from the frequency and sector concentration of the recent default events. Investors may not know who survives and who fails in the auto sector but the price they will charge for the perceived risk in anything autos could jump even from the already elevated levels we see today. exhibit 3
Delphi Ratings and Spreads
Rating
5yr CDS (bp) 0
BBB+/Baa1 13 BBB/Baa2 12 BBB-/Baa3 11
500
BB+/Ba1 10 BB/Ba2 9 BB-/Ba3 8
1,000
B+/B1 7 B/B2 6
1,500
B-/B3 5
2,000
S&P
CCC+/Caa1 4
Moody's
CCC/Caa2 3 CCC-/Caa3 2
5yr CDS (bp) 2,500 450
CC/Ca 1 Default 0
400
350
300
250 200 150 Days from Filing
100
50
0
Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
409
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 58
HY CDX INDEX ROLL – IMPLICATIONS
On a forward looking basis, the Dana bankruptcy is indeed one of the highlights of the upcoming roll into the High Yield CDX Series 6 (for which final constituents were recently announced). The high yield indices and tranches have actually been the most affected by the recent string of defaults. Most of the recently defaulted names were in all of the index series, largely because they occurred after the roll to Series 5; thus, even after these defaults, we have reasonably similar structures trading in the market for all the series. With the roll to CDX 6, we point out that the capital structure for this series will likely have upwards of 2% more subordination (or thickness in the case of the 0-10%) than the other series and a portfolio that is meaningfully different (10 names) from the current on-the-run HY CDX index. This should provide an interesting array of trading opportunities. Similarly, the removal of the auto names from the IG indices over time (and coincident addition to HY indices) should create similar opportunities. In Exhibit 4, we show the pricing we experienced for HY tranches during the last two index rolls. We note that because of meaningfully wider spreads in the new indices, as well as the extended maturities, we find that most tranches have traded meaningfully wider on the new on-the-run index. With the roll to Series 6, this effect may be more muted as the new portfolio does not trade meaningfully wider than the current (using recent levels) and the additional subordination in the new index serves to partially offset the impact of the increased maturity. Also in Exhibit 4, we hazard a guess as to where the new tranches could price on the roll. Our analysis is based on current market structure and past differences in implied correlation on roll dates. Clearly, these are first order estimates, but our goal is really to get a handle on how the subordination, portfolio and maturity differences are likely to affect pricing.
410
Please see additional important disclosures at the end of this report.
411
783 163
570 105 20
25-35%
35-100%
15
58
213
11%
4%
103
Diff
21
52
423
53%
83%
364
10/6/2005 CDX 4 4.7
Note: The differences for equity tranches are adjusted based on remaining notional after any losses. Source: Morgan Stanley
35
56%
46%
10-15%
15-25%
430 83%
327 78%
4/25/2005 CDX 4 5.2
0-10%
4/25/2005 CDX 3 4.9
40
113
693
65%
87%
433
10/6/2005 CDX 5 5.2
Where Are the New Tranches Going to Trade?
Index
Valuation Index Years to Maturity
exhibit 4
19
61
270
12%
9%
69
Diff
14
68
385
53%
83%
329
3/8/06 CDX 5 4.7
8
37 22
105
2% 117
502
56%
9 17%
338
Diff
87%
Estimate CDX 6 5.2
Section E
Performance and Ratings
Please see additional important disclosures at the end of this report.
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 59
CDO Performance – Unplugged
November 14, 2003
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Anisha Ambardar The significant transformation of the $400 billion CDO secondary market this year has, among other things, given structured credit investors an important source of performance information. In a market plagued for years by a lack of transparency, the liquidation of one of the world’s largest CDO senior note portfolios gave birth to secondary market liquidity, which has subsequently moved well beyond this one flow. From the moment we started researching the CDO market, we have been asked to estimate performance for the market. Without an active secondary market, hypothetical performance calculation was the only alternative. The results of hypothetical calculations tended to be somewhat amplified and far from pure (which we thankfully avoided). A SIMPLE SENIOR NOTE STUDY
Today, we feel confident enough to take a first look at CDO performance, although pricing information in the market is still somewhat fragmented. We therefore focus on a universe of over 100 floating rate senior notes for which we have monthly pricing information since inception (deals originating from 1995-2002). So how does it look? In a nutshell:
414
•
Senior notes have lost 1.3% on average per year, in excess of Libor. If the Libor component is included, the average performance would have been positive, but the focus in this study is on excess returns rather than on total returns.
•
Senior notes backed by leveraged loans have been the best performers (down less than 0.1% annually), while deals backed by high yield bonds were down 2.4% and investment grade cash bonds (synthetic not included) lost 3.0% on average.
•
Leveraged loan and emerging markets senior notes have exhibited the least return volatility (within the universe), while similar tranches from investment grade cash and high yield bond deals have experienced significantly more volatility.
•
2003 returns (year-to-date) have been dramatic: up 7.6% for the universe and 15.5% for high yield bond backed senior notes.
•
AAA senior notes have significantly outperformed AAs (-0.2% vs. -2.9%). Leveraged loan AAAs are the best performing at 0.5% annually.
•
Wrapped AAAs returned 0.2% on average. With high yield bond collateral, wrapped AAAs outperformed unwrapped AAAs (0.4% vs. -0.8%), but the opposite was true with leveraged loan collateral (0.2% vs. 0.5%).
•
1999 was the worst cohort for both AA- and AAA-rated senior notes from high yield deals. 1996 and 1997 were the best.
•
For the AAA-senior notes of leveraged loan transactions, 2000 was the worst cohort, while 2002 was the best. For AAs, 2001 was the worst and 1999 was the best.
A First Look at CDO Performance – Senior Notes: Excess Returns over Libor
exhibit 1
Emerging Markets
Since Inception -0.2%
2001 -0.4%
2002 -1.6%
2003 YTD 3.3%
Std Dev of Returns 0.5% 4.5%
Collateral High Yield Bonds
-2.4%
-3.6%
-16.2%
15.4%
US Leveraged Loans
-0.1%
0.1%
-3.6%
2.9%
1.7%
European Leveraged Loans
0.0%
0.1%
-1.1%
-1.7%
1.1%
Investment Grade (Cash)
-3.0%
-0.2%
-13.8%
5.2%
5.1%
Total (Senior Notes)
-1.3%
-1.9%
-10.6%
7.6%
3.6%
Source: Morgan Stanley
THE UNIVERSE AND METHODOLOGY
For this study, we included a total of 108 senior notes from deals that originated during the 1995 to 2002 period, where we have consistent monthly pricing information (see Exhibit 2). We included 43 high yield bond senior notes, 41 senior notes from leveraged loan transactions (including 12 from Europe), 12 investment grade cash senior notes, and eight from emerging markets. We excluded synthetic deals because the senior part of the capital structure is generally issued as a default swap, not a funded security. exhibit 2
Profile of Our Senior Note Sample
50
1995
1996
1997
1998
2000
2001
2002
European
43
1999
41
40
30
20 12
10
8
0 Emerging Markets
High Yield Bonds
Leveraged Loans
Inv. Grade Cash
Note: Four deals in the universe were categorized as “Other.” Source: Morgan Stanley
We calculated performance as excess return over Libor. Therefore, the return of each instrument during a month was based on percentage change of the flat price of the security, taking into consideration accrued interest from the excess coupon (above Libor) on the instrument. For example, a floating rate note with a Libor + 50 bp coupon priced at 99.0 with zero accrued at the end of month one and 99.5 at the end of month two would have an excess return of 0.55% for month two (equal to ((99.5 + (0.50/12))/99.0 -1). Aggregate returns are simple averages of individual tranche returns. Prices are mid-market.
Please see additional important disclosures at the end of this report.
415
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 59
THE PERFORMANCE BOOMERANG: 2002 VS. 2003
The sharp rally and compression of spreads in underlying credit markets during 2003 has had a resounding impact on CDO performance. Many floating rate senior notes trade at or near par (the median price in our universe is $97), which was simply unthinkable a year ago. Senior notes backed by high yield bonds had the most room to run, returning 15.5% this year, compared to being down 16.2% during 2002. The range of returns was from close to 0% to nearly 64% for these notes. Investment grade cash deals also had some room to run, and rallied over 5% in excess of Libor this year. Our limited universe in this sector was down nearly 14% in 2002, so the relative retracement has not been as strong as in high yield. Emerging markets senior notes (which are dominated by AA-rated securities) have rallied 3.3% above Libor this year, which is far less than high yield, even though the underlying markets have had a terrific run, but this can be explained by ratings (see below). In leveraged loans, 2003 returns are nearly 3%, while last year, the universe was down 3.5%. European leveraged loan senior notes are the only sub-sector that has generated negative returns during 2003, which we attribute to weaker secondary market liquidity, compared to US-originated transactions. As a matter of comparison, excess returns (above Treasuries) in investment grade corporate bonds this year have been close to 5%; in high yield, excess returns have been over 20%. AAA’S OUTPERFORM AA’S – WERE WRAPS NECESSARY?
The most important insight we gather from looking at the performance of senior notes across ratings is that the numbers reflect the fact that these structures are not necessarily free of impairment or stress. AA performance has lagged AAAs across the board since inception (-2.9% vs. -0.2%, based on original ratings). Within US leveraged loan deals, AAAs are up 0.5%, while AAs are down 0.4%. In both investment grade and high yield transactions, AAA notes are down about 0.8%. But the performance of AAs is uglier, with investment grade deals down 12.2% and high yield bond deals down 4.3%. Were wrapped securities necessary? Wrapped leveraged loan senior notes underperformed unwrapped AAAs (0.2% vs. 0.5%), but significantly outperformed in the high yield bond space (0.4% vs. -0.8%). AAAs Outperform AAs; Wraps Help High Yield But Not Leveraged Loans
exhibit 3
AAA Wrapped -0.2%
AAA
High Yield Bonds
0.4%
-0.8%
-4.3%
US Leveraged Loans
0.2%
0.5%
-0.4%
0.1%
-0.9%
Investment Grade (Cash)
-0.1%
-0.8%
-12.2%
Total
0.2%
-0.2%
-2.9%
Emerging Markets
European Leveraged Loans
Source: Morgan Stanley
416
AA -0.4%
CREDIT CYCLICALITY: A LOOK AT COHORTS
The cohort effect was most dramatic in high yield bond transactions, as one would expect, given the incredibly cyclical nature of default rates and the tendency for such deals to include new issuance. AAA and AA performance was in sync for high yield deals: 1999 was the worst year, while 1996 and 1997 were the best. exhibit 4
High Yield – Cohort Performance (Since Inception) Highlights Credit Cycles
2% 0% -2% -4% -6% -8% AAA
-10%
AA
-12% -14% -16% 1996
1997
1998
1999
2000
2001
Source: Morgan Stanley
exhibit 5
Leveraged Loans Cohorts
2%
AAA
AA
1%
1%
0%
-1%
-1% 1998
1999
2000
2001
2002
Source: Morgan Stanley
In the much more stable leveraged loan space, AAAs had positive performance since inception for all cohorts except 2000, while the performance of AAs reflects the credit cycle, albeit in a more muted way than in high yield. CONCLUSION – WHAT’S THE VALUE OF (NEGATIVE) CONVEXITY?
What are the trades to do? We argue that AAA- and AA-rated notes in most CDOs (backed by corporate credit) are negatively convex. Yet, that negative convexity may not be worth much to investors who are bullish on the credit markets over the medium term. We feel there is still value in many AA-rated senior notes for bullish investors.
Please see additional important disclosures at the end of this report.
417
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 60
Leverage and Long-Lasting Milk
January 16, 2004
Primary Analyst: Sivan Mahadevan Primary Analyst: Viktor Hjort Primary Analyst: Peter Polanskyj One of the biggest risks in levered long-structured credit strategies is idiosyncratic defaults, particularly when they happen quickly. Anyone involved in the synthetic CDO market during 2001 and 2002 has experience with this phenomenon. Two years forward, the character of this market has changed dramatically, but there is still idiosyncratic default risk out there. Given the dramatic improvements in pricing transparency, the impact of the Parmalat bankruptcy offers a lot of insight into the risks of structured credit investing in today’s markets. Parmalat is an idiosyncratic or “low correlation” event, meaning its credit issues had virtually no impact on the rest of the market. Protecting against such events is difficult unless the credits in question are individually hedged. Parmalat was in the Dow Jones Euro TRAC-X Index and was a common name in synthetic CDOs as well. Any investor who is long subordinate tranches is exposed to this idiosyncratic default risk to a fairly significant degree. Conventional wisdom dictates hedging some of this idiosyncratic risk, but hedging only a fraction of the risk is economically possible, and even then dynamic hedging strategies may not always work. We focus this chapter on describing both the ratings and mark-to-market impact of the Parmalat bankruptcy (as it unfolded) on structured credit, comparing actual pricing with what the models would have predicted. We also review how dynamic delta hedging strategies would have played out and offer some ideas to investors who seek to distinguish idiosyncratic default risk from market-wide risks. THE PARMALAT IMPACT ON BENCHMARK PRICING
Parmalat was a name that traded wide (200 bp) relative to the rest of the European investment grade market (40 bp) for most of 2003. In early November 2003, Parmalat spreads marched higher and stayed in the 500-600 bp range for two weeks before blowing out much further in mid-December. Parmalat filed for bankruptcy on December 24th. The initial 400 bp move in Parmalat had a 4 bp impact on the Dow Jones Euro TRACX portfolio (given its 1% contribution, although other credits tightened). The 0-3% tranche of the Dow Jones Euro TRAC-X moved 10 points (on an upfront basis) as a result of the initial move wider in Parmalat (see Exhibit 1). The 3-6% tranche widened 20 bp. The impact Parmalat had on more senior tranches was smaller than on the portfolio directly, reflecting the pivot point of leverage. The 9-12% tranche widened less than the 4 bp widening caused by Parmalat. Once Parmalat filed for bankruptcy, the 3-6% tranche widened an additional 80 bp (for a total of 100 bp) while the 9-12% tranche moved out an additional 2 bp, still outperforming the Dow Jones Euro TRAC-X portfolio. The overall loss on the 0-3% tranche is approximately 28.7%, with 26.7% attributed to the Parmalat default payments, and 2% attributed to repricing the (now thinner) equity tranche on the 99-name portfolio.
418
exhibit 1
The Parmalat Effect – Subordinate Tranches Feeling the Impact
TRAC-X, 3-6% and 9-12% spreads
450 TRAC-X (bid)
400
9-12%
3-6%
350 300 250 200 150 100 50 0 Jul-03
Aug-03
Sep-03
Oct-03
Nov-03
Dec-03
Note: Shows the spreads of Dow Jones TRAC-X Europe and tranches. Source: Morgan Stanley
In our view, the correlation moves on all tranches were not significant, which suggests that market behavior was consistent with the notion that it was an idiosyncratic event, with little impact on risk appetite going forward (see Exhibit 2). This effectively means that market pricing and expected pricing from the models should have been consistent. We discuss this further below. Yet, since the Parmalat bankruptcy, the implied correlation on the new 99-name European portfolio has been somewhat lower (for all tranches, see Exhibit 2). We attribute this to less demand for selling protection on subordinate tranches (which are long correlation trades) and more demand for selling protection on senior tranches (which are short correlation trades). This translates into greater demand for higher levels of subordination, which makes sense given increased fears of idiosyncratic default risk. exhibit 2
Correlation Falls After Parmalat
45% 40%
0-3%
Parmalat removed from Dow Jones Euro TRAC-X
6-9% 10-30%
35% 30% 25% 20% 15% 10% 5% Aug-03
Sep-03
Oct-03
Nov-03
Dec-03
Jan-04
Note: Shows the spreads of the Dow Jones TRAC-X Europe and implied correlation of tranches. Source: Morgan Stanley Research
Please see additional important disclosures at the end of this report.
419
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 60
TRADING STRATEGIES – TESTING THE MODELS
A risk in many of today’s structured credit strategies is that their performance deviates dramatically from model-predicted levels under adverse conditions. Risk managers and investors can lose a lot of sleep worrying about this. While there is no guarantee that the models will always work, model-based valuations were definitely in the right ballpark with respect to the Parmalat move and eventual default. As we noted above, we did not see big or unexplainable moves in correlation, which would generally be harbingers of modeling problems. Nevertheless, it is interesting to observe modelpredicted values (see Exhibit 3) and compare them to actual pricing. exhibit 3 Tranche 0-3% 3-6% 6-9% 9-12% 12-22%
Tranche Price Moves Relative to Underlying Portfolio
1 Credit Goes to 500 bp 22.5x 5.8x 1.4x 0.6x 0.1x
1 Credit Goes to 1,500 bp 22.1x 4.3x 0.9x 0.3x 0.1x
1 Credit Defaults (20% Recovery) 35.3x 3.7x 0.2x 0.4x 0.1x
Source: Morgan Stanley
As an example, the pricing model tells us that if a wide credit in a 40 bp portfolio widens to 500 bp, the 0-3% tranche should move 22.5 times the underlying portfolio. The move we observed in the 0-3% tranche was certainly of this magnitude: 7 bp on the portfolio times a spread duration of 4.5 equals a 0.31 price move on the Dow Jones Euro TRAC-X, which, if we multiply by 22.5 equals approximately 7 points. This is reasonably close to the initial move on the 0-3% tranche given that Parmalat actually widened beyond 500 bp in the period. For the 3-6% mezzanine note, the price move for a similar change in one credit is roughly 6 times the underlying portfolio, which again matches the initial move we saw as a result of Parmalat (4 bp * 6 = 24 bp). As a result of the default with recovery at 20%, the 3-6% should move approximately 4 times the underlying, according to the model. The 20 bp move on Dow Jones Euro TRAC-X (if the defaulted Parmalat is included) was a factor of 5 smaller than the 100 bp move on the 3-6% tranche. DYNAMIC DELTA HEDGING
For those who protected their subordinate tranches (and many did) against a Parmalat default, the deltas that come out of these models are fairly important, as they provide guidance regarding how much single-name protection to purchase. We show the approximate singlename deltas for an average credit in Dow Jones TRAC-X NA in Exhibit 4. Assuming a $30 million 0-3% tranche exposure, it is interesting to observe that an average credit has a delta of $4.2 million (14%). The delta rises to $8.7 million (29%) if the credit becomes distressed. The maximum exposure to a single name is $10 million (33%). For mezzanine and more senior tranches, the deltas tend to fall as a single credit becomes distressed (indicating a low i-gamma) because of the increased probability of that credit defaulting and impacting only the most subordinate tranche (which has a high i-gamma), leaving the mezzanine and senior tranches with less subordination to a lower yielding portfolio.
420
exhibit 4
Deltas to an Average Credit Getting Wider
0-3%
Initial Delta 14%
Delta with Credit at 500 bp 23%
Delta with Credit at 1000 bp 29%
3-7%
6%
5%
4%
7-10%
1%
1%
0%
Source: Morgan Stanley
FUNDED NOTES – WATCH OUT FOR RATINGS ACTIONS
Parmalat is a common credit in synthetic CDOs. Fitch estimates that EUR 700 million of Parmalat exposure is included in the 69 deals it has rated. Moody’s has rated 65 CDOs with approximately EUR 600 million of Parmalat exposure in total (there is overlap in these two universes). On December 24, Fitch put 29 securities from 11 public CDOs and an additional 24 tranches from 22 private CDOs on ratings watch negative. Parmalat exposure in these transactions ranged from 0.20% of the underlying collateral to an astonishing 5%. Moody’s has found portfolio exposure to be between 0.13% and 1.33% for synthetic CDOs, and between 1.2% and an even more astonishing 6.3% for four cash flow CDOs it has rated. Fitch’s actions, at that time, were based on expectations of a 40% recovery rate, but with bonds trading closer to 20% now, there could be more severe ratings actions by that agency. To date, Fitch has downgraded one CDO transaction; Moody’s has downgraded several. We expect many more downgrades as the agencies plow through all of the deals in their respective universes. Synthetic CDO equity tranches, which are generally not rated even if they are issued as funded securities, were largely held by the dealer community and prop desks, based on our experiences. Given that Parmalat was always a wide-trading credit, we believe that most dealers dynamically delta hedged Parmalat prior to the events, and certainly during the march toward bankruptcy. Dynamic delta hedging does not guarantee the avoidance of loss though, as such a strategy is highly dependent on recovery rate views and the cost of protection purchased. Given expectations for a low recovery, we do believe that there will be losses in broker-dealer correlation books, but those losses may not be significant. As for hedge funds, we would not be surprised to see a significant loss somewhere, owing to equity or levered mezzanine tranches. POSITIONING FOR IDIOSYNCRATIC DEFAULTS?
The recent spread widening of Adecco (a name in the Dow Jones Euro TRAC-X Index which recently announced that it expects the audit of its financial statements for 2003 to be delayed) is a reminder that more idiosyncratic credit risk exists in the markets. How can investors position for such events? We have recommended buying protection on senior tranches (10-15% for example) as a hedge against market-wide moves in spreads. But, as the Parmalat episode has shown, such tranches move less than the underlying market when single-name events occur. Individual “bolts from the blue” have a much bigger impact on more subordinate tranches. For investors looking to play such events, buying protection on mezzanine-type tranches is one way to implement the view, but we caution that such trades have large negative carry and can suffer significantly if spreads tighten. The trade only makes sense for those with a conviction that more idiosyncratic events are to come.
Please see additional important disclosures at the end of this report.
421
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 61
Revisiting CDO Ratings Migration
March 17, 2006
Primary Analyst: Vishwanath Tirupattur Primary Analyst: Sivan Mahadevan The degree to which CDO ratings migrations are reflective of the “true” credit quality of CDO tranches is arguable. Nevertheless, it is undeniable that ratings actions have a profound impact on credit markets. One needs to look no further than the changes to index inclusion rules of popular fixed income indices and the auto sector downgrades of 2005 to appreciate the hold that ratings have on the credit investor mindset. For cash CDO tranches, the significance of ratings as a metric for tranche credit quality is even more pertinent, given their complex structural features and the relatively lower levels of secondary market activity. With this in mind, we analyzed a large database of Moody’s CDO ratings actions 1 in a report we published last year (“Understanding Asymmetry in CDO Ratings Migration,” February 11, 2005) to provide a useful historical perspective to CDO performance over time. We reported a clear asymmetry in CDO ratings actions in terms of downgrades and upgrades and a substantial degree of variability in ratings actions over time and across different vintages and CDO types. In this chapter, we update the analysis by extending the data to the end of 2005. DATA
The data now encompasses more than 2,000 ratings actions over the period 1997-2005 – a significantly large timeframe, covering the troughs and peaks of a full credit cycle. As was the case in last year’s study, the database of ratings actions is predominantly US CDOs, which clearly dominate the universe of outstanding CDO tranches. As before, we treated pari passu tranches as one tranche to avoid manifold counting of ratings actions in CDOs with multiple tranches with the same ratings. We did not include withdrawn ratings 2 or insured tranches. Further, in a slight departure from the last year’s study, we excluded arbitrage synthetic CDO rating actions from the analysis. This was because we felt that the database’s coverage of arbitrage synthetic CDOs was not comprehensive. Consequently, this analysis now focuses on cash CDOs. With aforementioned adjustments, we have a total of 1,596 ratings actions in the dataset, consisting of 1,490 downgrades and 106 upgrades. We used the ratings that were prevalent immediately preceding the ratings action to determine if it is a downgrade or an upgrade, as opposed to using the ratings at the beginning of the year as in some other comparable studies. Where there have been multiple ratings actions in a given year in one tranche, we have counted each action separately and have not treated them as one as in some ratings agency studies. Finally, while the results
1
Monthly updates of “CDO Ratings Changes – Excel Data Supplement” obtained from Moody’s Investors Service web site. 2 Moody’s generally considers withdrawn ratings to be an indicator of a strong credit performance since most withdrawn ratings are for tranches that have received full payment of principal and all accrued interest.
422
reported here are in terms of number of tranches, we did the same analysis in terms of initial dollar par volume. The results are qualitatively similar. 2005: A VERY GOOD YEAR
Ratings actions during 2005 accounted for 48.1% of all CDO upgrades and only 6.9% of all CDO downgrades during 1997-2005 in our dataset. While downgrade actions still significantly outnumber upgrades for CDOs taking the ratings actions over the entire period into consideration, 3 the addition of another year’s ratings actions resulted in the downgrade-upgrade ratio improving from about 24 downgrades to every upgrade to 14 downgrades to every upgrade. For 2005, the ratio was about 2 to 1. ABS CDOs of 20002002 vintages accounted for nearly 90% of all downgrades during the year, while investment grade and high yield CBOs accounted for the majority of the upgrades. Ratings agency studies indicate that when evaluated as a single sector, CDO downgrades are similar to those of corporate securities. Based one-year transition rates, downgrades in CDO tranches are roughly comparable to corporate securities in general (Exhibit 1). exhibit 1 Rating Aaa
Average One-Year Downgrade Risk All CDOs 4.7%
Corporates 3.6%
Aa2
12.0%
11.4%
Baa2
14.7%
13.8%
Baa3
18.5%
12.7%
Ba2
18.2%
20.2%
Ba3
17.7%
18.7%
Source: Moody's and Morgan Stanley Calculations
However, we note a significant variability in terms of CDO type, vintage year and initial ratings, which are detailed in the following sections. The figures in Exhibits 2 – 6 are expressed as a percentage of total downgrades or upgrades in our sample data of ratings actions.
3
We have discussed the reasons for the high downgrade/upgrade ratio in our report, “Understanding Asymmetry in CDO Ratings Migration,” February 11, 2005.
Please see additional important disclosures at the end of this report.
423
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 61
VINTAGE EFFECTS
There is a clear vintage effect in CDO ratings actions. CDOs issued between 1997-2001 account for more than 80% of all CDO downgrade actions and two-thirds of all upgrades. It is worth pointing out that of the 1,490 downgrades in our data, more than 1,000 took place just in the two years between 2002-03. As we have noted before, corporate credit loss rates peaked during 2001-02, while CDO downgrade activity peaked during 2002-03, suggesting a lag between timing of credit losses in the underlying sector and their manifestation in CDO ratings actions. While this reaction time has been coming down, CDO ratings actions still lag developments in the underlying collateral sectors. The 1999 vintage of deals (predominantly high yield CBOs) stands out in that it has contributed to most downgrades (in 2002) and also upgrades (in 2005). Moody’s CDO Rating Actions by Vintage
exhibit 2 (% of Total Downgrades) Vintage Year 1990
Grand Total
Year of Rating Action 1998 0.0%
1999 0.0%
2000 0.0%
2001 0.0%
2002 0.0%
2003 0.0%
2004 0.0%
2005 0.0%
0.1%
1995
0.0%
0.0%
0.0%
0.0%
0.2%
0.0%
0.0%
0.0%
0.2%
1996
0.1%
0.3%
0.1%
0.5%
2.3%
0.5%
0.1%
0.0%
4.0%
1997
0.9%
0.8%
1.7%
2.3%
4.4%
1.3%
0.6%
0.1%
12.1%
1998
0.2%
0.6%
1.1%
3.6%
8.5%
1.9%
0.7%
0.3%
16.9% 26.7%
1999
0.0%
0.0%
0.0%
2.0%
17.0%
6.4%
1.1%
0.1%
2000
0.0%
0.0%
0.0%
0.7%
4.6%
10.9%
4.6%
1.3%
22.1%
2001
0.0%
0.0%
0.0%
0.1%
3.9%
4.9%
2.9%
2.6%
14.4%
2002
0.0%
0.0%
0.0%
0.0%
0.1%
0.2%
0.7%
2.3%
3.4%
2004
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.1%
0.1%
1.2%
1.7%
3.0%
9.2%
41.1%
26.2%
10.7%
6.9%
100%
Grand Total
(% of Total Upgrades) Vintage Year 1996
1997 0.9%
2000 0.0%
2001 0.0%
2002 0.9%
2003 0.9%
2004 0.0%
2005 2.8%
5.7%
1997
0.0%
1.9%
2.8%
4.7%
0.9%
1.9%
5.7%
17.9%
1998
0.0%
0.0%
0.0%
1.9%
0.9%
4.7%
6.6%
14.2% 23.6%
1999
0.0%
0.0%
0.9%
1.9%
0.9%
4.7%
15.1%
2000
0.0%
0.0%
0.0%
0.0%
5.7%
1.9%
5.7%
13.2%
2001
0.0%
0.0%
0.0%
0.0%
11.3%
1.9%
1.9%
15.1%
2002
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
3.8%
3.8%
2003
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
6.6%
6.6%
Grand
0.9%
1.9%
3.8%
9.4%
20.8%
15.1%
48.1%
100%
Source: Moody's and Morgan Stanley Calculations
424
Grand Total
Year of Rating Action
DIVERGING PERFORMANCE: CLOS, CBOS AND ABS CDOS
There is wide variation in ratings performance across CDO types. CLOs, which account for over 30% of all outstanding cash CDO tranches account for less than 6% of all CDO downgrades over the 1997-2005. While they account for a higher proportion of all CDO upgrades (13.2%), it is still significantly smaller than their share of the overall outstanding CDO tranches. This underscores the strength of CLO ratings stability through the vicissitudes of a credit cycle. This ratings stability of CLOs is despite the recent pick-up in default rates, including several high profile issuers. As we have argued ad nauseam, this ratings stability hinges critically on high post-default loan recoveries experienced in nearly all of the defaults experienced during 2005. In a latter section, we discuss our thoughts on the prospects for this ratings stability to continue over the next credit cycle. Moody’s CDO Rating Actions by CDO Type
exhibit 3 (% of Total Downgrades) CDO Type HY Bond CBO
1998 0.4%
1999 1.1%
Year of Rating Action 2000 2001 2002 2003 2.4% 6.0% 32.1% 15.0%
2004 3.2%
2005 0.4%
Grand Total 60.5% 6.0%
IG Bond CBO
0.0%
0.0%
0.0%
0.0%
1.7%
3.8%
0.2%
0.3%
Leveraged Loan CLO
0.0%
0.3%
0.3%
0.5%
1.7%
2.5%
0.1%
0.2%
5.8%
EM CDO
0.7%
0.0%
0.2%
0.3%
0.3%
0.0%
0.1%
0.0%
1.7%
ABS CDO
0.0%
0.0%
0.0%
0.0%
0.3%
3.0%
6.7%
6.0%
16.0%
Balance Sheet CDO
0.1%
0.3%
0.1%
2.3%
3.8%
1.3%
0.4%
0.0%
8.3%
Other CDO
0.0%
0.0%
0.1%
0.1%
1.0%
0.7%
0.0%
0.0%
1.8%
Grand Total
1.2%
1.7%
3.0%
9.2% 41.1% 26.2% 10.7%
6.9%
100%
(% of Total Upgrades) CDO Type HY Bond CBO
1997 0.9%
2000 0.0%
Year of Rating Action 2001 2002 2003 0.0% 0.9% 3.8%
2004 11.3%
2005 26.4%
Grand Total 43.4%
IG Bond CBO
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
1.9%
1.9%
Leveraged Loan CLO
0.0%
0.0%
0.0%
0.0%
4.7%
2.8%
5.7%
13.2%
Middle Market CLO
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
2.8%
2.8%
EM CDO
0.0%
1.9%
2.8%
5.7%
0.9%
0.0%
0.0%
11.3% 14.2%
ABS CDO
0.0%
0.0%
0.9%
0.0%
1.9%
0.0%
11.3%
Balance Sheet CDO
0.0%
0.0%
0.0%
2.8%
6.6%
0.0%
0.0%
9.4%
Other CDO
0.0%
0.0%
0.0%
0.0%
2.8%
0.9%
0.0%
3.8%
Source: Moody's and Morgan Stanley Calculations
Please see additional important disclosures at the end of this report.
425
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 61
On an overall basis, high yield CBOs were the source of the majority of downgrades accounting for over 60% of all CDO downgrade actions. However, during 2005, they accounted for a mere 6% of CDO downgrades but more than 50% of CDO upgrades. This is not entirely surprising, given that corporate credit defaults while on the up-tick in the latter half of 2005, remained quite low by historical standards, certainly a lot lower than the 2001-02 levels. In addition, most of the vintage high yield and investment grade CBOs of yesteryear have been delevering. Coupled with a still benign corporate default environment during 2005, delevering has stabilized the coverage tests for the many remaining CBO tranches. Much of the turmoil in corporate credit during 2005 pertained to autos and auto parts sectors. Recall that new issuance of high yield CBOs effectively ended in 2002. Consequently, during the period when the high yield CBOs were acquiring assets, many of the auto sector issuers currently undergoing various degrees of credit turmoil were rated investment grade, and, as such, high yield CBO portfolios had rather limited exposure to them. As a result, the 2005 auto sector turbulence has left the high yield CBOs relatively unscathed. The ratings performance of ABS CDOs follows an entirely different story line. ABS CDOs contributed to 16% of all CDO downgrades. While downgrades in other CDO types have been on a declining trend, they have been increasing in ABS CDOs. It is important to note though that there is a clear vintage effect to the growing downgrades in ABS CDOs. Nearly all of them are from vintages of 2002 and prior. As noted in Chapter 54, this is entirely attributable to the differences in the collateral composition of ABS CDOs of 2002 and prior vintages and subsequent vintages. In Exhibit 4, we summarize the ratings actions by the initial rating of a CDO tranches. CDO tranches with an initial rating of Baa2/Baa3 have experienced most downgrades, accounting for 37% of all downgrades, followed by tranches with an initial Aa2 rating (14.6%), which also have highest proportion of upgrades.
426
Moody’s CDO Rating Actions by Initial Rating
exhibit 4 (% of Total Downgrades) Initial Rating 1998 1999 Aaa 0.0% 0.0%
2000 0.0%
0.3%
2004 1.9%
2005 1.1%
Grand Total 11.0%
0.1%
0.2%
1.9%
Aa1
0.0%
0.0%
Aa2
0.2%
0.3%
Aa3
0.1%
0.0%
A1
0.0%
0.0%
A2
0.2%
0.1%
A3
0.0%
0.0%
Baa1
0.0%
0.1%
Baa2
0.0%
0.0%
Baa3
0.5%
0.9%
Ba1
0.0%
0.0%
0.0%
0.1%
0.3%
0.1%
0.0%
0.0%
0.5%
Ba2
0.0%
0.1%
0.3%
1.1%
3.5%
2.2%
0.5%
0.3%
7.9%
Ba3
0.0%
0.2%
0.0%
0.4%
2.4%
2.1%
0.5%
0.4%
6.0%
B1
0.1%
0.1%
0.3%
1.1%
1.9%
0.2%
0.0%
0.0%
3.8%
B2
0.0%
0.0%
0.0%
0.1%
0.2%
0.0%
0.0%
0.0%
0.3%
B3
0.0%
0.0%
0.1%
0.3%
0.7%
0.1%
0.0%
0.0%
1.3%
Grand Total
1.2%
1.7%
3.0%
9.2%
41.1%
26.2%
10.7%
6.9%
100%
(% of Total Upgrades) Initial Rating 1997 2000 Aaa 0.0% 0.0%
0.1%
Year of Rating Action 2001 2002 2003 0.3% 4.1% 3.5% 0.1%
1.0%
0.7%
1.3%
5.4%
3.4%
2.0%
1.4%
14.6%
0.1%
0.1%
2.3%
0.7%
0.3%
0.5%
4.2%
0.0%
0.1%
0.1%
0.1%
0.1%
0.1%
0.6%
0.0%
0.3%
1.3%
0.7%
0.3%
0.2%
3.0%
0.1%
0.2%
1.9%
3.0%
1.0%
0.1%
6.2%
0.0%
0.1%
0.7%
0.1%
0.1%
0.1%
1.3%
0.2%
1.2%
8.5%
7.5%
3.1%
1.6%
22.1%
1.2%
2.2%
6.7%
2.1%
0.9%
0.9%
15.4%
Year of Rating Action 2001 2002 2003 0.0% 0.0% 1.9%
2004 2.8%
2005 8.5%
Grand Total 13.2%
Aa1
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
3.8%
3.8%
Aa2
0.9%
0.9%
0.9%
3.8%
2.8%
5.7%
8.5%
23.6%
Aa3
0.0%
0.0%
0.0%
0.0%
0.9%
0.9%
3.8%
5.7%
A1
0.0%
0.0%
0.0%
0.0%
1.9%
0.0%
2.8%
4.7%
A2
0.0%
0.0%
0.9%
3.8%
0.9%
0.9%
2.8%
9.4%
A3
0.0%
0.0%
0.0%
0.0%
1.9%
0.9%
2.8%
5.7%
Baa1
0.0%
0.0%
0.0%
0.0%
1.9%
0.9%
0.0%
2.8%
Baa2
0.0%
0.0%
0.0%
0.9%
1.9%
1.9%
5.7%
10.4%
Baa3
0.0%
0.9%
0.9%
0.9%
1.9%
0.9%
6.6%
12.3%
Ba1
0.0%
0.0%
0.0%
0.0%
3.8%
0.0%
0.0%
3.8%
Ba2
0.0%
0.0%
0.9%
0.0%
0.9%
0.0%
0.9%
2.8%
Ba3
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.9%
0.9%
B2
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.9%
0.9%
Grand Total
0.9%
1.9%
3.8%
9.4%
20.8%
15.1%
48.1%
100%
Source: Moody's and Morgan Stanley Calculations
Please see additional important disclosures at the end of this report.
427
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 61
Exhibits 5 and 6 focus on downgrade severity. In Exhibit 5, we show the average number of notches a tranche has been downgraded across different CDO types, given its initial rating. In Exhibit 6, we show the most a tranche has been downgraded to, given its original rating across different CDO types. exhibit 5
Average Number of Notches Downgraded
Initial Rating
Aaa
CDO TYPE ABS CDO 2.8
HY Bond Leveraged CBO Loan CLO 2.5
Aa1
2.6
2.6
Aa2
3.7
2.8
1.4
Aa3
4.0
2.6
1.3
A1
3.5
A2
5.4
3.7
A3
4.0
2.7
IG Bond CBO 2.0
EM CDO
3.0
2.0
3.3
2.8
3.5
Other CDO 2.1
1.0 2.0
2.9
2.0
2.9
6.0 4.0
Balance Sheet CDO 1.6
2.5
2.2 2.3
2.9
3.0
1.0
Baa1
4.0
2.8
Baa2
4.6
3.1
2.5
3.9
4.5
2.7
3.3
Baa3
4.2
3.1
3.2
3.7
2.8
3.0
2.0
Ba1
3.3
2.0
3.0
8.0
3.8
Ba2
4.4
3.8
2.9
1.0
2.6
Ba3
5.1
4.0
4.1
3.6
3.0
2.7
Source: Moody's and Morgan Stanley Calculations
The average downgrade severity is modest in CLOs and most severe in ABS CDOs. It is worth pointing out that in our dataset, no CLO tranches originally rated Aaa have been downgraded.
428
exhibit 6
Average Number of Notches Downgraded
Initial Rating
CDO TYPE ABS CDO
HY Bond CBO
Aaa
Baa2
A2
Aa1
A3
A3
Leveraged Loan CLO
Aa2
Ba2
Ba1
A2
Aa3
Ba2
Baa2
A2
A1
Baa2
A2
Ba3
Ba3
A3
B3
Caa1
A3
IG Bond CBO
Balance Sheet CDO
Other CDO
A1
A2
Aa3
A1
Aa2
Baa3
EM CDO
A1
Baa2
A2
Ba2
Ba1 Baa3
Ba1
Ba2
A3
Baa3 Baa2
Ba2
Ba1
Baa1
Baa1
B1
Ba3
Baa2
Caa3
Caa1
Ba3
Caa3
B3
Caa1
B1
Baa3
B3
Ca
Caa2
B3
B2
Ba3
Ba3
Ba1
Ba3
B2
Ba3
Caa3
Caa2
Ba2
C
Ca
B3
Ba3
Caa3
Ba3
C
Ca
C
Ca
B3
B3
Source: Moody's and Morgan Stanley Calculations
WHERE TO NEXT?
CDO ratings follow the credit performance of the underlying sectors, albeit with a potential lag. Therefore, for post-2002 ABS CDOs, future ratings action will follow those of US housing and the RMBS sector. For reasons discussed earlier, mainly delevering, we expect the recent positive momentum in vintage CBOs transactions to continue in the remainder of the year, barring of course, a sudden spike in corporate defaults which would be outside our baseline expectations for corporate defaults. While we remain constructive about the continued ratings stability of CLOs in the near term, we temper our sanguinity with some caution. Two points we have noted elsewhere (See “Who Puts the “L” in LBO and More,” February 17, 2006) drive our caution. Issuer diversification across multiple CLO tranches is somewhat limited, more so in US CLOs. Therefore, a spike in defaults has the potential to be far more widespread and thus damaging in the next down turn in the credit cycle whenever it happens. Second, while loan recoveries have been high in issuers that recently defaulted, we must keep in mind that these were companies with debt capital structures dominated by unsecured bonds versus secured loans. In several high profile LBOs, we have noticed that secured loans are a rising proportion of the debt capital structure. If future corporate defaults arise from firms with loan-heavy capital structure, prospects for recoveries would be dimmer. While the CLO ratings impact of the current Moody’s proposal to supplement ratings with probability of default and loss given default ratings remains unclear and widely debated, one possibility that has been speculated is that managers of existing CLOs might respond to the changes by increasing their utilization of second lien and bond buckets to trade off spread versus recovery. The application of such strategies may not be constructive for continued CLO ratings stability. We acknowledge the contribution of Simmi Sareen to this chapter.
Please see additional important disclosures at the end of this report.
429
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 62
Interest Rates – Rising Tides Lift CDO Boats?
March 4, 2005
Primary Analyst: Vishwanath Tirupattur Primary Analyst: Sivan Mahadevan INTEREST RATE WORRIES
US Treasury markets in the past few weeks have been roiled by remarks from Federal Reserve officials, conflicting signals from a flurry of economic statistics and growing concerns about foreign central bank appetite for USD assets. The Treasury yield curve has experienced significant flattening. Ten-year Treasury rates have been down by more than 30 bp and have come back up more than 40 bp since the beginning of the year. No less authority than the Chairman of Federal Reserve, Alan Greenspan, characterized the recent rally in Treasury bonds as a “conundrum.” The surprising jump in the producer price index certainly rekindled inflation fears, which were never entirely dormant. Crude oil prices, which had dipped to below $42 per barrel toward the end of December 2004, have roared back to above $53 per barrel and have continued to stoke inflation fears. Morgan Stanley economist Richard Berner points to the dwindling slack in the economy, rising costs combined with slower productivity growth and a stillaccommodative monetary policy as the three factors contributing to upside risks to future inflation. He suggests that the Fed requires a significantly tighter monetary policy in order to realize its forecasts of stable core inflation through 2006. (See “The Next Inflation Surprise,” February 22, 2005.) Since June 2004, the Fed has already raised the target Fed Funds rate six times, 25 bp each time, for a sum of 1.50% (Exhibit 1). The Morgan Stanley economics team expects another 25 bp increase in each of the next two upcoming Fed meetings. Fed Funds futures and Eurodollar contracts seem to be expecting another 150 bp hike by June 2006. While there are differences of opinion regarding the amount of monetary tightening to come over the next year, it seems beyond question that the short-term interest rates are headed higher.
430
exhibit 1
Federal Funds Target Rate
2.6% 2.4% 2.2% 2.0% 1.8% 1.6% 1.4% 1.2%
Jan-05
Feb-05
Dec-04
Oct-04
Nov-04
Sep-04
Jul-04
Aug-04
Jun-04
Apr-04
May-04
Mar-04
Jan-04
Feb-04
1.0%
Source: Morgan Stanley
Given the specter of rising short-term interest rates, what are the implications for CDO performance? Prima facie, it would seem that since CDO debt is predominantly of floating rate nature, rising interest rates are of little consequence. However, we suggest that interest rates do affect the performance of certain types and have an even more prominent impact in certain vintages of CDOs. Depending upon the extent of monetary tightening, the impact can be meaningful in some cases. An examination of the effect of interest rate movements on CDO performance is the focus of this chapter. MINIMAL IMPACT ON CLO PERFORMANCE
CDOs whose assets are also largely floating rate instruments are mostly unaffected by interest rate movements except for the indirect macro economic effects such movements may have on default probabilities of the underlying assets and their recoveries upon default. The level of interest rates also may have other indirect effects through refinancing and prepayments within the CDO collateral. At current interest rate levels and taking into consideration the range of expected changes, it seems to us that this is a second order effect and likely to have minimal impact on CDOs with predominantly floating rate collateral. CLOs largely fall into this category and, as such, are relatively insulated from interest rate changes. POTENTIAL EFFECTS ON CBOs AND ABS CDOs
However, in the case of high yield or investment grade CBOs and ABS CDOs, the underlying collateral is dominated by fixed rate instruments, and the liabilities have a substantial floating rate component. The potential interest rate mismatch is addressed through interest rate hedges at deal inception, in the form of interest rate swaps and/or interest rate caps. Of these two hedging instruments, interest rate swaps are the most significant component of the hedge in the vast majority of CDOs with a hedge in place at deal inception. The swaps are generally structured as amortizing swaps – swaps whose notional balance declines according to a predetermined schedule. The amortization schedule is determined at deal inception to match the expected amortization schedules of the CDO’s debt tranches. Coupons from the fixed rate bonds pay the fixed rate leg of the swap and the floating rate receipts from the swap offset the coupons on the floating rate debt tranches. Exhibit 2 demonstrates how such an interest rate swap hedge works in a typical CDO.
Please see additional important disclosures at the end of this report.
431
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 62
exhibit 2 Interest Rate Swap Counterparty
Typical Interest Rate Swap Hedge
Rec Floating
Pay Fixed
Rec Fixed
Pay Floating Floating Rate Debt Tranches
Fixed Rate Assets
CDO Rec Floating
Pay Fixed
Floating Rate Assets Residual
Fixed Rate Debt Tranches Equity
Source: Morgan Stanley
In a small number of cases, balance-guaranteed swaps are used with the amortization of the swap notional explicitly linked to that of the CDO’s debt tranches. In other words, the swap notional amortizes as a function of the outstanding notional of the CDO’s floating rate debt tranches, rather than according to a predetermined schedule. Since such swaps are significantly more expensive than the plain-vanilla interest rate swaps with pre-determined fixed amortization schedules, the latter have been the hedge instruments in the vast majority of CDOs. Interest caps, when employed as hedges, are structured to be deep-out-of-the-money protection against rising interest rates. Typically, they are structured at deal inception to be forward starting. It is important to note that payments to hedge counterparty are usually at the top of the cash flow waterfall for most CDOs, ahead of payments to the senior-most debt tranches. HEDGE AS A SOURCE OF RISK?
During the life of a CDO, a deal can become over-hedged or under-hedged depending upon the realized performance of CDO assets and/or liabilities relative to the expectations at deal inception. As fixed rate assets default, the source of cash flow for the fixed rate payments to the hedge counterparty diminishes. Thus, when interest rates decline under an over-hedge scenario, the deal experiences a drag on its cash flows, having to pay a higher fixed rate to the hedge counterparty while receiving lower floating rates. Failure of performance tests in terms of over-collateralization, coverage ratios and collateral credit quality can result in the diversion of available cash flows towards faster-than-scheduled amortization of the senior debt tranches, which results in realizing a liability payoff schedule different from that envisaged at deal inception, based on which swap amortization schedules would have been determined. The extent of this hedge mismatch determines a CDO’s sensitivity to interest rate movements. The higher a CDO’s actual performance deviates from its expected performance, the higher the potential for a hedge mismatch, and consequently, the higher the sensitivity to interest rate movements.
432
DEFINING HEDGE MISMATCHES
It is important to remember that if the realized deal performance is close to the expected performance at deal inception, there will be little hedge mismatch; thus, interest rate movements will not have a direct impact on CDO performance. Our understanding of how rating agencies estimate the over-hedge amount is the following: Subtract the sum of the performing floating rate assets from the sum of the current par amount of the floating rate debt tranches of the CDO to compute the required amount of the hedge notional. The notional amount of the swap available in the deal in excess of the required hedge notional is the amount by which a deal is considered currently over-hedged. We illustrate using the example of PPM America High Yield CBO I Limited, a high yield bond CBO that closed in March 1999. Fitch downgraded the A-1 notes from BBB- to BB on February 23, 2005, citing, among other reasons, the over-hedged nature of the deal. Based on data reported by Intex, the current par amount of the outstanding floating rate tranches is about $147 million. The collateral portfolio has $140 million of assets, of which about 99% are fixed rate assets and 1% floating rate assets. The transaction has a current outstanding swap notional of $244 million. 1 This results in an over-hedged notional amount of about $98.6 million and constitutes the source of the deal’s sensitivity to interest rate movements. It must be noted that this measure of the hedge mismatch does not take into account the amortization of the swap. A more refined approach would incorporate not only the amortization schedule of the swap but also current expectations of the CDO’s asset defaults and pay downs of the CDO’s debt tranches. AS RATES RISE, OVER-HEDGE HELPS
Exhibit 3 shows the 3-month LIBOR rates over the last several years. During 2002-03 when interest rates declined, transactions that originated during 1998-2000 experienced significant drag on their performance. Poor collateral credit performance in terms of defaults was further exacerbated in those deals by the drag on the deal cash flows from the underlying hedges. As interest rates rise, this effect is reversed. Over-hedged deals stand to benefit in terms of cash flow and the mark-to-market value of the hedges. With rising short-term rates, the spread between the fixed rate payments to and the floating rate receipts from the hedge counterparty declines, cushioning the deal cash flow. The mark-to-market value of the hedge also improves, thus improving the market value and liquidation over-collateralization ratios, deal parameters of significance in the secondary markets for CDO tranches. The higher the short-term rates are, the more pronounced their beneficial impact on over-hedged CDOs. It is precisely the over-hedged deals of this vintage (1998-2000) that bore the brunt of the effect of falling short-term rates that stand to benefit from rising rates. The longer the remaining term to maturity of the swap, the higher the sensitivity of deals with hedge mismatches to interest rate changes.
1
The swap starts to amortize in June 2005 based on a predetermined schedule, declining to a zero balance by December 2008 (Source: Intex).
Please see additional important disclosures at the end of this report.
433
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 62
exhibit 3
Three-Month LIBOR (1998-present)
8.0% 7.0% 6.0% 5.0% 4.0% 3.0% 2.0% 1.0%
Jul-04
Jan-05
Jul-03
Jan-04
Jan-03
Jul-02
Jan-02
Jul-01
Jan-01
Jul-00
Jan-00
Jul-99
Jan-99
Jul-98
Jan-98
0.0%
Source: Morgan Stanley
We illustrate the potential mark-to-market impact of the hedge mismatch with an example using the PPM America High Yield CBO I referred to earlier. As a part of the hedge, the deal pays a fixed interest rate of 5.45% and receives 6-month LIBOR. The current outstanding swap notional is $244 million, which amortizes to zero by the end of 2008. Exhibit 4 shows the mark-to-market impact of interest rate movements on this interest rate swap by shifting the forward curve by 50 bp, 100 bp, 150 bp and 200 bp. exhibit 4
Interest Rate Sensitivity of Typical Hedges in CBOs
Current
Mark-to-Market Value ($9,029,633)
Forward curve + 50 bp
($6,694,595)
Forward curve + 100 bp
($4,358,827)
Forward curve + 150 bp
($2,023,059)
Forward curve + 200 bp
$312,708
Source: Morgan Stanley, Company Reports
The mark-to-market value of the hedge is currently negative to the tune of about $9 million. As rates rise, the drag on the deal from the hedge declines. An upward shift of 200 bp in the forward curve moves the hedge into positive territory. The shape of the curve also matters. In a bear flattener scenario (short-term rates rising more than longterm rates), the mark-to-market value of an over-hedge improves even more. We do not want to overstate the effect of rising short-term interest rates on CDO performance and valuation. All we are suggesting is that CDOs that suffered because of the over-hedge problem when short-term rates declined are likely to benefit when short-term rates rise. The beneficial impact depends upon the extent of the hedge mismatch and remaining term to maturity of the swap.
434
RATING ACTIONS FROM OVER-HEDGE
It is worth emphasizing that exposure to interest rate movements through a hedge mismatch is secondary to the credit quality of the collateral and realized default rates in determining CDO performance. Still, the drag on the deal cash flows arising out of hedge mismatches was one reason cited as rationale for the many CDO downgrades experienced during that period. Exhibit 5 shows a list of CDO rating actions in which hedge mismatches were cited as contributing factors for downgrade actions. While by no means meant to be comprehensive, this list of 64 unique CDO tranches illustrates that hedge mismatch is a significant issue considered by rating agencies. This list can also serve as a starting point for identifying transactions that would most benefit from rising short-term interest rates.
Please see additional important disclosures at the end of this report.
435
436 ACF HY Bond
Eastman Hill Funding I Ltd
Eastman Hill Funding I Ltd
Eastman Hill Funding I Ltd
Eastman Hill Funding I Ltd
7/2/2003
7/2/2003
7/2/2003
7/2/2003
South Street CBO 1999-1
South Street CBO 1999-1
South Street CBO 1999-1
3/30/2003
3/30/2003
South Street CBO 1999-1
ACF HY Bond
ML CBO VII 1997-C-3
3/30/2003
ACF HY Bond
J/Z CBO (Delaware)
2/4/2003
2/11/2003
3/30/2003
ACF HY Bond
J/Z CBO (Delaware)
J/Z CBO (Delaware)
2/4/2003
2/4/2003
ACF HY Bond
ACF HY Bond
ACF HY Bond
ACF HY Bond
ACF HY Bond
ACF HY Bond
ACF HY Bond
ACF HY Bond
ACF HY Bond
ACF HY Bond
Summit CBO I Ltd
Summit CBO I Ltd
ACF HY Bond
10/25/2002
Summit CBO I Ltd
10/25/2002
ACF HY Bond
ACF HY Bond
ACF HY Bond
ACF HY Bond
ACF HY Bond
CDO Type ACF HY Bond
10/25/2002
ML CBO VII 1997-C-3
ML CBO III 1996-C-1
3/13/2002
ML CBO VI 1996-C-2
3/13/2002
3/13/2002
Summit CBO I Ltd
Summit CBO I Ltd
9/21/2001
9/21/2001
Deal Summit CBO I Ltd
07/02/2001
07/02/2001
07/02/2001
07/02/2001
06/03/1999
06/03/1999
06/03/1999
06/03/1999
03/13/1997
05/16/2000
05/16/2000
05/16/2000
04/23/1999
04/23/1999
04/23/1999
08/29/1996
03/13/1997
10/31/1996
04/23/1999
04/23/1999
Deal Closing Date 04/23/1999
A-3
A-2 IO
A-1 FX
A-1 FL
A-2
A-2L
A-1
A-1LB
A
C
B
A
D-1
C
B
A
A
A
D-1
C
10
522
10
512
36
24
55
10
170
19
22
108
6
36
37
184
170
154
6
36
AA
AAA
AAA
AAA
AAA
AAA
AAA
AAA
AA
BBB
A
AAA
BB
BBB
AA-
AA-
AA
AA
BB
BBB
AA
AAA
AAA
AAA
CCC
CCC
A
A
CCC+
BBB
A
AAA
CC
CC
BB
A-
BBB
BBB-
BB
BBB
BBB
AA
AA
AA
CCC-
CCC-
BBB
BBB
CC
B+
BBB
AA-
C
C
C
BB
CCC+
CCC+
CC
CC
Original Original Previous Revised Principal Rating Rating Rating ($MM) Currency Class B 37 AAAABB
Select CDO Downgrade Actions Due to Over-Hedge Exposure
BBB
AA
AA
AA
CC
CC
BBB-
BBB-
CC
CC
B+
AA-
C
C
C
BB
CC
D
C
C
Current Rating C
Fitch
Fitch
Fitch
Fitch
S&P
S&P
S&P
S&P
S&P
S&P
S&P
S&P
Fitch
Fitch
Fitch
S&P
S&P
S&P
Fitch
Fitch
Agency Fitch
chapter 62
Date 9/21/2001
exhibit 5
Structured Credit Insights – Instruments, Valuation and Strategies
Please see additional important disclosures at the end of this report.
437
01/24/2001 01/24/2001
Triton CDO IV Ltd
MWAM CBO Series 2001-1 Ltd ACF SF
MWAM CBO Series 2001-1 Ltd ACF SF
MWAM CBO Series 2001-1 Ltd ACF SF
9/10/2003
9/10/2003
9/10/2003
Phoenix CDO II Ltd
Mid Ocean CBO-2000-1 Ltd
Mid Ocean CBO-2000-1 Ltd
Mid Ocean CBO-2000-1 Ltd
Mid Ocean CBO-2000-1 Ltd
Ingress I, Ltd
Ingress I, Ltd
E*TRADE ABS CDO I, Ltd
11/11/2003
4/30/2004
4/30/2004
4/30/2004
4/30/2004
7/12/2004
7/12/2004
8/11/2004
Sutter CBO 1999-1, Ltd
Sutter CBO 1999-1, Ltd
10/15/2003
10/15/2003
Sutter CBO 1999-1, Ltd
Sutter CBO 1999-1, Ltd
10/15/2003
10/15/2003
ACF SF
ACF SF
ACF SF
ACF SF
ACF SF
ACF SF
ACF SF
ACF SF
ACF HY Bond
ACF HY Bond
ACF HY Bond
ACF HY Bond
ACF HY Bond
MWAM CBO Series 2001-1 Ltd ACF SF
Sutter CBO 1999-1, Ltd
9/10/2003
10/15/2003
ACF HY Bond
07/02/2001
Combo
10/08/2002
05/18/2000
05/18/2000
01/08/2001
01/08/2001
01/08/2001
01/08/2001
05/17/2000
11/17/1999
11/17/1999
11/17/1999
11/17/1999
11/17/1999
01/24/2001
01/24/2001
12/15/1999
B
C
B
B-1
A-2L
A-2
A-1L
B
B2
B1-L
B1-L
A4-L
A4
Pref Shr
C-2
C-1
B
B
25
21
54
13
15
17
240
39
10
5
6
5
24
8
13
22
198
27
18
USD
USD
USD
AA+
A-
AA
BBB-
AA-
AA-
AAA
AA
BB-
BBB
BBB
A-
A-
BB-
BBB
BBB
AA
AA-
B+
BBB
AA+
BB-
A+
B
A-
A-
BBB
AA
BB-
BBB
BBB
A-
A-
BB-
BBB
BBB
AA
BB+
B+
BBB
A-
B-
A-
CCC
BB
BB
BBB
BBB
BB-
BBB
BBB
A-
A-
B-
BB-
BB-
A-
B
C
CC
Original Original Previous Revised Principal Rating Rating Rating ($MM) Currency Class B-1 24 BBB BBB B
07/02/2001 Sub Pref
7/30/2003
ACF HY Bond ACF HY Bond
Eastman Hill Funding I Ltd
Eastman Hill Funding I Ltd
7/2/2003
CDO Type ACF HY Bond
Deal Closing Date 07/02/2001
Select CDO Downgrade Actions Due to Over-Hedge Exposure
7/2/2003
Deal Eastman Hill Funding I Ltd
Date 7/2/2003
exhibit 5 (cont.)
A-
B-
A-
CCC
BB
BB
BBB
BBB
B+
BB+
BB+
BBB+
BBB+
B-
BB-
BB-
A-
CCC-
C
CC
Current Rating B
Fitch
Fitch
Fitch
Fitch
Fitch
Fitch
Fitch
Fitch
Fitch
Fitch
Fitch
Fitch
Fitch
Fitch
Fitch
Fitch
Fitch
S&P
Fitch
Fitch
Agency Fitch
438 ACF SF ACF SF ACF SF
8/11/2004 E*TRADE ABS CDO I, Ltd
8/11/2004 E*TRADE ABS CDO I, Ltd
8/11/2004 E*TRADE ABS CDO I, Ltd
ACF HY Bond 07/13/2000 ACF HY Bond 07/13/2000
9/10/2004 Wilbraham CBO Ltd
9/10/2004 Wilbraham CBO Ltd
ACF SF ACF SF ACF SF
12/6/2004 Bleecker Structured Asset Funding, Limited
12/6/2004 Bleecker Structured Asset Funding, Limited
12/6/2004 Bleecker Structured Asset Funding, Limited
ACF SF
ACF SF
8/25/2004 Bleecker Structured Asset Funding, Limited
ACF SF
ACF SF
9/27/2004 Varick Structured Asset Fund Ltd
03/29/2000
ACF SF
8/25/2004 Bleecker Structured Asset Funding, Limited
8/25/2004 Bleecker Structured Asset Funding, Limited
9/27/2004 Varick Structured Asset Fund Ltd
03/29/2000
ACF SF
8/25/2004 Bleecker Structured Asset Funding, Limited
02/08/2001
03/29/2000
03/29/2000
03/29/2000
09/29/2000
09/29/2000
03/29/2000
03/29/2000
02/08/2001
ACF SF ACF SF
8/18/2004 MKP CBO I, Limited
02/08/2001
02/08/2001
03/29/2000
03/29/2000
10/08/2002
8/18/2004 MKP CBO I, Limited
ACF SF ACF SF
8/18/2004 MKP CBO I, Limited
8/12/2004 Bleecker Structured Asset Funding Ltd.
8/18/2004 MKP CBO I, Limited
ACF SF ACF SF
8/12/2004 Bleecker Structured Asset Funding Ltd.
C-2
B
A-2
A-1
A-2
A-1
A-2
A-1
C
B
A-2
A-1
B-1L
B-1A
A-2L
A-1L
A-2
A-1
Pref Shr
10/08/2002 Comp Sec
10/08/2002
CDO Type ACF SF
Deal Date 8/11/2004 E*TRADE ABS CDO I, Ltd
40
315
45
233
39
19
141
34
40
315
45
7
7
25
250
131
19
13
5
3
USD
USD
USD
USD
USD
USD
USD
USD
USD
USD
USD
USD
USD
USD
USD
USD
USD
USD
USD
USD
AA
AAA
AAA
AAA
AAA
AA
AAA
BBB
AA
AAA
AAA
BBB-
BBB-
AA-
AAA
AAA
AAA
BBB-
BBB-
BBB
B-
A-
A-
A-
A-
BBB-
A+
B
BBB
AAA
AAA
BBB-
BBB-
A
AAA
AA
AA
BBB-
BBB-
BBB
CC
BBB
BBB
BB+
BB+
B
BBB+
C
B-
A-
A-
B+
B+
BBB-
AA-
A+
A+
CCC-
CCC-
B+
CC
BBB
BBB
BB+
BB+
B
BBB+
C
CC
BBB
BBB
B-
B-
BB
A-
A+
A+
CCC-
CCC-
B+
Fitch
Fitch
Fitch
S&P
S&P
S&P
S&P
Fitch
Fitch
Fitch
Fitch
Fitch
Fitch
Fitch
Fitch
S&P
S&P
Fitch
Fitch
Fitch
Original Original Previous Revised Current Principal Class ($MM) Currency Rating Rating Rating Rating Agency C-1 10 USD BBB BBB B+ B+ Fitch
Select CDO Downgrade Actions Due to Over-Hedge Exposure
chapter 62
Deal Closing Date 10/08/2002
exhibit 5 (cont.)
Structured Credit Insights – Instruments, Valuation and Strategies
Please see additional important disclosures at the end of this report.
439
10/11/2002 10/11/2002 10/11/2002 10/11/2002
ACF SF ACF SF ACF SF ACF SF ACF SF ACF SF
12/6/2004 Varick Structured Asset Fund, Ltd
12/6/2004 Varick Structured Asset Fund, Ltd
12/6/2004 Varick Structured Asset Fund, Ltd
1/24/2005 Bristol CDO I Ltd
1/24/2005 Bristol CDO I Ltd
1/24/2005 Bristol CDO I Ltd
ACF HY Bond 11/17/1999
2/28/2005 FC CBO III Ltd
ACF= Arbitrage cash flow; SF=Structured Finance; HY = High Yield Source: Morgan Stanley, S&P, Fitch
ACF SF ACF HY Bond 03/02/1999
1/24/2005 Bristol CDO I Ltd
2/23/2005 PPM America High Yield CBO I
09/29/2000
09/29/2000
09/29/2000
CDO Type ACF SF
Deal Closing Date 09/29/2000
B
A-1
C
B
A-2
A-1
B-2
B-1
A-2
39
449
13
30
21
224
7
25
300
USD
USD
USD
USD
USD
A-
AAA
BBB
AA
AAA
AAA
A-
A-
AAA
CCC
BBB-
B
A+
AAA
AAA
CCC-
CCC-
BBB-
CCC-
BB
C
A+
AA+
AA+
CC
CC
BB
CCC-
BB
C
A+
AA+
AA+
NR
NR
BB+
S&P
Fitch
Fitch
Fitch
Fitch
Fitch
Fitch
Fitch
Fitch
Original Original Previous Revised Current Principal Class ($MM) Currency Rating Rating Rating Rating Agency A-1 50 USD AAA BBBBB BB+ Fitch
Select CDO Downgrade Actions Due to Over-Hedge Exposure
Deal Date 12/6/2004 Varick Structured Asset Fund, Ltd
exhibit 5 (cont.)
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 63
Correlation or Credit – What Drives Ratings Changes?
March 4, 2005
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Ajit Kumar, CFA In the early days of the synthetic CDO market, ratings were the main drivers of valuation and relative value, and most investors were quite focused on rating agency methodologies and their evolution. Today we have a lot more experience and information with respect to valuation and pricing sensitivity of synthetic CDO tranches, due in part to the explosion of activity and liquidity, industry standard correlation models, and the benefit of having been through a rather sharp credit cycle. In many ways, the markets have moved well beyond the rating agencies, which not too surprisingly is causing a bit of a scramble at the agencies. Ratings, and their perceived stability, nevertheless remain important considerations for a large number of investors. While the agencies themselves established their CDO ratings models years ago, there has been a flurry of methodology changes recently, spurred by the increasing transparency in and growing sophistication of investors. The current battle among the agencies involves correlation, or to be more specific, the correlation assumptions that go into their models. In our view, the agencies are trying to solve a problem that they have created, given that they use industry-standard models in non-standard ways. In contrast, today’s risk-neutral correlation models show that correlation has a second order effect on the tranches that are most likely to be rated. Who is right? In our opinion, it does not matter, because changes in correlation have not been drivers of ratings instability. The primary driver of ratings changes in synthetic CDOs over the years has been changes in credit quality of the underlying portfolio – in other words, credit selection. This may sound like an obvious statement, but there is an important message here. An underlying CDO portfolio with the same average rating as the market does not necessarily perform exactly like the market. Portfolio sampling risk is the most important driver of CDO ratings changes relative to the ratings shifts of the underlying market. The larger the portfolio, the smaller the sampling risk; but many of yesterday’s synthetic CDOs did not have large underlying portfolios, so sampling risk remained a big issue. Today the market is somewhat different, with many portfolios indeed much larger, particularly CDO-squared structures.
440
THE RATING AGENCY DEBATE
Reacting to a market that has moved ahead of them, the agencies have now adopted approaches based on correlation models to rate synthetic CDOs (including CDOsquareds). The three agencies’ approaches differ somewhat though, as do their correlation assumptions, and this has been the source of a public debate recently. Market-standard correlation models are run in a risk-neutral sense, meaning that the probability of default for a given credit is derived from market pricing. For example, a credit that trades at 50 bp (assuming flat curve) has about a 0.83% chance of defaulting every year, assuming a 40% recovery. Given full credit curves for all credits in a CDO’s portfolio, the market-standard models can relatively easily calculate prices given a correlation assumption, or alternatively, imply correlation from market prices. The point is not that the default probabilities are absolutely correct but that they provide a benchmark from which to price the tranches, which reflects the market perception of absolute and relative risk within the portfolio given the maturity of the transactions. exhibit 1
Correlation Sensitivity in Different Spread Environments
Mark-to-Market Impact of 1% Change in Correlation
0-3%
1.0%
3-7% 7-10% 10-15%
0.8%
15-30% 0.6% 0.4% 0.2% 0.0% -0.2% 0
25
50
75 100 125 Index Spread Level (bps)
150
175
200
Source: Morgan Stanley
The agencies now use similar models, but not in a risk-neutral sense because their methodologies are not attempting to value instruments but rather to assign some real world probability of default. They derive a credit’s default probability from ratingsbased historical averages. This process implies three very important outcomes. First, the portfolios feel much less risky than those that use market-based default probabilities, because they exclude market risk premiums. Second, the resulting portfolios are significantly more homogenous than portfolios segmented by spread, with much thinner tails. Third, the risk in the portfolios appears more stable through time as compared to volatility of market spreads. These portfolios, combined with agency stress factors, tend to be very sensitive to correlation for certain types of junior and senior mezzanine tranches in the agencies’ models (hence the debate).
Please see additional important disclosures at the end of this report.
441
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 63
exhibit 2
Correlation Sensitivity is Low Today for Mezzanine
Impact on Market Value Corr. + 1% Corr. + 10% DJ CDX Ser 3 IG - 5 Yr
Equivalent Index Move in bp Corr. + 1% Corr. + 10%
0-3%
0.61%
6.10%
-0.8
-7.8
3-7%
-0.07%
-0.67%
0.2
2.2
7-10%
-0.07%
-0.75%
0.6
6.2
10-15%
-0.02%
-0.19%
0.3
3.9
15-30%
-0.02%
-0.20%
1.0
10.4
DJ CDX Ser 3 IG - 5 Yr Bespoke Tranches 1-3%
0.46%
4.26%
-0.5
-5.0
2-4%
0.05%
0.83%
-0.1
-1.4
3-5%
-0.09%
-0.46%
0.2
1.2
4-6%
-0.12%
-0.82%
0.4
3.0
5-7%
-0.11%
-0.85%
0.6
4.3
6-8%
-0.10%
-0.81%
0.7
5.2
7-9%
-0.09%
-0.75%
0.7
5.8
8-10%
-0.08%
-0.68%
0.8
6.4
0-3%
0.84%
8.41%
-1.6
-15.7
3-7%
0.41%
3.92%
-0.7
-6.2
7-10%
-0.03%
-0.06%
0.1
0.2
10-15%
-0.04%
-0.38%
0.2
2.2
15-30%
-0.04%
-0.36%
0.5
4.8
0-10%
0.43%
4.19%
-5.1
-49.1
10-15%
0.20%
1.92%
-1.4
-12.9
15-25%
-0.12%
-0.92%
1.1
8.6
25-35%
-0.11%
-1.00%
2.7
23.3
DJ CDX Ser 3 IG - 10 Yr
DJ CDX Ser 3 HY - 5 Yr
Source: Morgan Stanley
CORRELATION IS A SECOND ORDER EFFECT TODAY
To confirm our assertion above, we have calculated correlation sensitivities for various mezzanine tranches of investment grade and high yield portfolios in Exhibit 1, based on the market’s benchmark instruments and models. The sensitivity of a tranche’s price to correlation varies with the size and seniority of tranches, the level of spreads, and the shape of the distribution of the portfolio, among other factors. However, the tranches that are the least sensitive to correlation today are the ones that are most likely to be rated by the agencies (everything except the highly subordinated and super-senior). We do need to put these sensitivities into perspective. In Exhibit 2 we show the implied price change of a parallel shift in correlation. We also calculate the equivalent spread move on the underlying index to achieve the same price change. The average weekly standard deviation of the Dow Jones CDX 3 index has been 1.9 bp since it was launched last year. That being said, changes in spread can obviously be a more important driver in values – and, by extension, risk – than correlation.
442
WHAT DRIVES RATINGS SHIFTS?
If correlation is not that important, what should impact ratings changes? As we pour through ratings changes data for synthetic CDOs over the years, we note that most activity is driven by shifts in credit quality (up or down) of the underlying portfolios and actual defaults. While the investment grade market itself has had quite a bit of such activity over the past 5 years, we believe that most CDOs that have experienced significant negative ratings migration have had portfolios that have underperformed the market (again, this may sound obvious). Actual defaults aside, this phenomenon is a double-edged sword, though, as many of these same deals have experienced upgrades recently (31 upgrades this year; see Exhibit 3). exhibit 3
Downgrades
Ratings Migration of Synthetic CDOs and Corporate Bonds 2001 1
2002 42
Upgrades A-Baa Realized/Average Downgrades
2003 33
2004 2
2005 (YTD) 1
3
1
31
111%
157%
127%
68%
72%
46%
59%
130%
Ba-B Realized/Average Downgrades
144%
171%
133%
91%
Ba-B Realized/Average Upgrades
115%
71%
95%
219%
A-Baa Realized/Average Upgrades
Note: 2001-2004 data include only Moody's rating actions. 2005 includes rating actions from Moody's, S&P and Fitch for January and February. Note: Realized/Average transition is defined as the calendar year cohort transition divided by the 19832004 average transition. Source: Morgan Stanley
PORTFOLIO SAMPLING RISK
Portfolio sampling risk can be modeled generically and has been used historically to help design optimal basket sizes (for our early thoughts, see “Understanding Mezzanine Notes,” June 24, 2002 and Chapter 12). While actually quantifying the impact of portfolio size on portfolio performance versus some broader universe is difficult, several useful insights can be gleaned from simple statistics. In Exhibit 4, we show the uncertainty of expected losses for portfolios of varying sizes. We can think of sampling risk as the difference in uncertainty (standard deviation) between a portfolio with a given number of credits and the uncertainty of a portfolio containing the entire universe, which we assume is 1,000 credits below. In Exhibit 5, we show this differential for portfolios of varying sizes all with expected losses of around 3.2% over 5 years.
Please see additional important disclosures at the end of this report.
443
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 63
exhibit 4
Uncertainty of Expected Losses
Standard Deviation of Expected Losses 4.0% Independent
20% Correlated
3.5% 3.0% 2.5% 2.0% 1.5% 1.0% 0.5%
950
1000
900
850
800
750
700
650
600
550
500
450
400
350
300
250
200
150
50
100
0.0%
Portfolio Size
Source: Morgan Stanley
We have performed the study under two assumptions. The first is that default risk among the portfolio components is independent. The sampling risk in this portfolio drops off very quickly and the volatility of the losses levels off to a reasonably low absolute level (0.50% or 16% of the expected losses) as the portfolio size increases to include the entire universe. Under the second assumption, we add positive correlation into the mix, which has the effect of increasing the uncertainty (volatility) of the expected losses to a much higher level. This is true not only for small portfolios but also for large portfolios (standard deviation is 3.1% or nearly 100% of the expected losses). Sampling risk drops off quickly but uncertainty of actual default experience for the entire universe increases dramatically. This uncertainty in the loss experience, even for a portfolio containing the entire universe, can be thought of as the main driver of relative portfolio performance. exhibit 5
Portfolio Sampling Risk Falls as the Number of Credits Increases
1.6% 1.4%
Independent
20% Correlated
1.2% 1.0% 0.8% 0.6% 0.4% 0.2%
Portfolio Size
Source: Morgan Stanley
444
1000
950
900
850
800
750
700
650
600
550
500
450
400
350
300
250
200
150
50
100
0.0%
Another way to see this phenomenon is through the historical default statistics. If actual correlation were very low, then the default statistics would be reasonably stable over time. We studied the volatilities of the Moody’s cumulative 5-year default rates and found them to be 160%, 150% and 70% of the average default rates for the Aa, A, and Baa rating buckets, respectively. The results indicate that the broad universe actually experiences quite volatile default rates, similar in order of magnitude to our correlated scenario above. LARGE PORTFOLIOS ARE MORE STABLE FOR SENIOR NOTES
Perhaps we have used a long-winded approach to make a very simple point, but the analysis is always important to support the thesis. For senior tranche investors who are seeking high ratings and relative ratings stability in synthetic tranches, credit selection is a key driver, until portfolios get large enough that the likelihood of deviating from the averages is relatively small. For senior note investors who do not want to worry about credit selection, today’s large pool CDO-squared structures (200 to 300 names adjusted for overlap) may provide the relative stability at the senior level, despite all of the correlation chatter at the agencies.
Please see additional important disclosures at the end of this report.
445
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 64
Looking Behind Curtains
September 16, 2005
Primary Analyst: Vishwanath Tirupattur Primary Analyst: Sivan Mahadevan It would be a huge understatement to say that ratings matter to cash CDO investors. Notwithstanding the oft-leveled criticism about methodological biases, subjectivity and timeliness issues, CDO rating actions remain the most widely understood and accepted gauge as a performance metric and a measure of relative value on an ongoing basis. The complex waterfall mechanisms, coverage and collateral quality tests, embedded interest rate hedges and delevering features of CDOs imply that a number of factors can influence the performance – and hence the rating – of a CDO tranche. When Moody’s, S&P and Fitch announce a rating change, the rationale for the action is described in some detail in their press releases. While there are differences between rating agencies in terms of the details of the rationales that are spelt out, the trend has been towards more openness and clarity of the reasons for the rating actions. Understanding the rationale and the frequency with which each different reason is cited for a rating agency action provides a useful insight into the relative importance of the many factors that affect CDO ratings. It may serve as a guide for CDO investors in determining the factors to focus on in their surveillance and analysis of their CDO investments. Excluding withdrawn and affirmed ratings and watch/outlook changes and treating rating actions on pari passu tranches as one rating action, there have been 961 CDO rating changes over the last 14 months (July 2004–August 2005) by the three agencies combined. Based on the publicly available press releases, we have identified the reasons for each of these ratings actions. We have categorized the reasons for the rating actions into a number of categories separately for downgrades and upgrades (Exhibit 1). We have then analyzed them by aggregating them by CDO type and vintage. Some of these reasons are clearly interrelated. To the extent more than one reason has been noted in a rating action, we have counted them as such. We have also aggregated them by rating agency to see if there are any discernible differences between agencies as well. We do recognize that each agency’s actions are a function of the portfolio of transactions rated by them and some of the differences between agencies are a consequence of that alone. Still, it would be interesting to see if there are differences.
446
exhibit 1
Classification of Ratings Rationale
Upgrades U1 Transaction restructuring resulting in payments in excess of original structure U2
Improvements in coverage tests
U3
Seasoning/Improved credit quality of reference portfolio
U4
Amortization/delevering of capital structure
U5
Upgrade of put counterparty and/or liquidity put provider; additional guarantees
Downgrades D1 Deterioration of collateral quality D2
Failure of coverage tests
D3
Distressed securities in collateral pool, recovery expected to be low
D4
Technical default and consequent acceleration of senior tranches (negative for subordinated tranches)
D5
Downgrade of put counterparty and/or liquidity put provider
D6
Exposure to MH/airline lease assets
D7
Negative impact of interest rate hedge
Source: Morgan Stanley
DELEVERING AND COLLATERAL QUALITY MATTER MOST
On an overall basis, seasoning and improved credit quality of the underlying portfolio and amortization/delevering emerge as the most significant reasons cited by agencies for CDO upgrades. An improvement in coverage tests, which is a consequence of delevering, is cited less frequently. For downgrades, deterioration of collateral quality is cited as a rationale in an overwhelming number of cases, more than twice as frequently as the next closest cited rationale of failing coverage tests. Distressed securities with low expected recoveries is the next most frequently cited reason for downgrades, often in conjunction with the credit quality deterioration, which can be thought of as rating agencies considering “jump-to-default” risk. In a number of cases, accelerated amortization of senior tranches was cited as the reason for the downgrade of a subordinated tranche, often in conjunction with the upgrade of the senior tranche. In Exhibit 2, we summarize the rationale behind rating actions by type of CDOs. There are significant differences between different CDO types. Considering that a vast majority of the HY bond CBOs have begun to delever, it is not surprising that amortization and/or delevering is the most cited reason for their upgrades, much more so than the improvement in credit quality. It may be noted that many of these transactions had experienced significant credit quality deterioration in the last downturn in credit cycle, and subsequent improvement in the collateral credit quality is not as material as delevering and improvement of coverage tests. To some extent, this is true for all arbitrage cash CDOs. On the other hand, for synthetic CDOs with relatively less stringent cash flow diversion mechanisms, improvement in credit quality is the dominant reason for upgrades.
Please see additional important disclosures at the end of this report.
447
448 4 8
SF CDO
Synthetic CDO
Market Value CDO
Balance Sheet CDO 160
31
7
4
57
-
21
38
270
58
6
176
5
-
5
18
Upgrade Rationale U2 U3 2 2
Source: Morgan Stanley, Moody’s, S&P and Fitch
20
-
IG Bond CBO
Total
5 3
HY Loan CLO
U1 -
203
33
4
10
28
2
38
87
U4 1
8
-
-
1
7
-
-
-
U5 -
357
9
1
38
263
8
-
38
D1 -
149
-
3
-
121
4
1
20
D2 -
106
4
4
14
67
5
1
11
40
-
2
1
26
-
2
9
Downgrade Rationale D3 D4 -
Ratings Rationale by CDO Type (July 2004 – August 2005)
8
-
-
5
3
-
-
-
D5 -
54
-
-
4
50
-
-
-
D6 -
44
-
-
-
40
-
1
3
D7 -
chapter 64
HY Bond CBO
CDO Type EM CDO
exhibit 2
Structured Credit Insights – Instruments, Valuation and Strategies
For downgrades, deteriorating collateral quality has been paramount for SF CDOs, which has experienced the most downgrades of all CDO categories during the period covered by this study. This deterioration is mostly attributable to exposure to troubled structured finance sectors such as manufactured housing, cited explicitly in 50 rating actions. The manufactured housing sector also dominates the distressed securities. It is also significant that the negative impact of interest rate hedges has been cited only with respect to SF CDOs. A generally benign corporate credit environment has characterized the period covered by this study. As such, downgrades in corporate credit backed CDOs have been relatively few. Still, it is worth noting that credit quality deterioration has been the most cited reason for their downgrades, more material than other reasons, including the failing coverage tests. VINTAGE EFFECTS
Exhibit 3 summarizes the rationale by vintage. The difference in the significance of different reasons across CDO types is consistent with their performance to date. Not surprisingly, delevering is more material for older vintage CDOs. Given that a vast majority of downgrades has been in SF CDOs, it is useful to note a vintage effect in the rationale. Deteriorating collateral quality is the predominant reason for 2000-2002 vintages only. It may be recalled that SF CDOs of these vintages were multi-sector CDOs. The change in the nature of underlying collateral into US real estate securitizations clearly accounts for this significant vintage effect.
Please see additional important disclosures at the end of this report.
449
U1 1 1 5 6 4 3 20
450 U1 6 5 9 20
Upgrade Rationale U2 U3 17 95 42 83 101 92 160 270
Source: Morgan Stanley, Moody’s, S&P and Fitch
Agency 0 S&P Fitch Total
exhibit 4
U4 1 1 5 16 31 46 25 16 18 43 1 203
U5 1 1 6 8
D1 1 2 11 19 72 145 87 8 9 3 357
D2 1 4 5 6 36 68 28 1 149
Downgrade Rationale D3 D4 1 2 10 5 4 6 27 10 45 11 17 6 1 1 106 40
U4 40 80 83 203
U5 1 6 1 8
D1 146 95 116 357
D2 76 16 57 149
Downgrade Rationale D3 D4 22 7 25 10 59 23 106 40
Ratings Rationale by Rating Agency (July 2004 – August 2005)
Upgrade Rationale U2 U3 5 6 17 12 39 9 15 30 11 51 34 78 39 70 14 160 270
Ratings Rationale by CDO Vintage (July 2004 – August 2005)
D5 2 6 0 8
D5 1 2 5 8
D6 7 19 28 54
D6 8 20 24 1 1 54
D7 0 8 36 44
D7 2 19 12 11 44
chapter 64
Source: Morgan Stanley, Moody’s, S&P and Fitch
Vintage 1988 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 Total
exhibit 3
Structured Credit Insights – Instruments, Valuation and Strategies
DIFFERENT STROKES FOR DIFFERENT FOLKS?
We summarize the rationale for CDO rating actions by rating agency. As noted earlier, we recognize that rating actions and their cited rationale are a function of the portfolio of transactions rated by each agency and the surveillance mechanisms in place at each agency. For the period under study, Fitch had the most upgrade actions and Moody’s had the most downgrade actions. Improvement in coverage tests is cited more frequently by Fitch, relative to Moody’s and S&P, for upgrade actions. Similarly, the negative impact of interest rate hedges is cited a lot more frequently by Fitch than by the other agencies. CONCLUSION
We have peered behind the curtains to analyze the rationale offered by the rating agencies for CDO rating actions. It is worth repeating that the period under study in this chapter is not representative of a full credit cycle. Still, the study provides insights to the rationale used by rating agencies for CDO rating actions. Collateral quality and delevering seem to be the most cited reasons for rating actions. There are significant differences between cash and synthetic CDOs in terms of the rationale for rating actions. While some difference between rating agencies is understandable as being reflective of the transactions rated by them and their differing take on the underlying collateral, questions linger in our mind about consistency of rating actions and timeliness of surveillance. We acknowledge the contributions of Simmi Sareen to this chapter.
Please see additional important disclosures at the end of this report.
451
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 65
Tranche Performance – Mezz Sits Tight & October 28, 2005 Equity Takes a Ride Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj While the term “performance attribution” does not necessarily roll off anyone’s tongue, understanding how one actually earned the performance they received is fairly important, particularly as the calendar year-end approaches. Clearly much has happened in structured credit markets over the past two quarters, and there will be some big performance winners and losers this year on both shores of the buy- and sellside divide. Directionally, investment index equity tranches are net down in total return since the GM-induced auto-stress event started in the spring, while mezzanine and senior tranches have posted solid positive performances, although the ride has been very bumpy. But the attribution is equally important, and, in the case of equity, has been associated with spread widening and dispersion, as well as adverse correlation moves, which more than offset any benefit from carry and time decay. For mezzanine tranches, any underlying spread widening was offset by changes in implied correlation. The high yield tranches have earned their performance in a similar manner, although aggregate spread levels played a larger role and the Collins & Aikman’s default impacted the 0-10% tranche performance. The 15-25% tranche showed very strong performance, driven by implied correlation changes, spread impact and carry. REVISITING 2004 AND 1Q05
As those who read our report on tranche performance in the first quarter of 2005 (see “Zooming in on Tranche Performance,” April 15, 2005) may recall, our methodology attributes the performance of tranches to seven factors: changes in average spreads, change in average curve shape, parallel correlation shifts, changes in correlation skew, changes in spread dispersion, time decay and, finally, carry. While the relative importance of these factors depends on one’s time horizon, clear patterns have emerged. In our last report on this topic, we pointed out that correlation was a key driver on the positive side for 2004 mezzanine returns, regardless of maturity or underlying. This, coupled with the broad-based spread tightening in the period, led to significant positive performance across the capital structure. In the first quarter of this year, despite rhetoric about systemic risk being on the rise, mezzanine and senior tranches continued to outperform the underlying index as spreads widened, though they did experience negative absolute returns. The period we cover with this performance study began in April shortly before the news of Kerkorian’s tender for GM’s shares and ended, quite ironically, just before Delphi’s recent bankruptcy filing. Given the many ups and downs, we find it enlightening to examine the drivers of tranche performance in the last two quarters, in particular. It is no surprise that correlation continued to be a key driver of relative performance.
452
IG TRANCHE HOBBITS – THERE AND BACK AGAIN
In Exhibit 2, we show the attribution of tranche performance from the beginning of April through the end of September of this year for the IG and HY CDX 4 index tranches (the YTD numbers assume a roll from CDX 3 in April). One of the key takeaways for the period is that the performance of investment grade equity tranches was directionally different than mezzanine tranches in the context of index spreads that are only marginally wider, in aggregate. Despite all the action, at the end of September, the 5-year index sat within 5 bp of the level at the beginning of April. Even for CDX 3, which contains Delphi, these dynamics were generally true, with 5- and 10-year equity positions losing between 10% and 13%, while the 5-year 3-7% tranche experienced a rally similar to that of the 3-7% tranche of CDX 4 (these tranches recently traded almost flat to one another). exhibit 1
Tail Risk in CDX 4 Amplified by Implied Correlations
0-3% Correlation 0%
Widest Credits Rel. to Index 6x Widest as % of Index 5x
5%
Equity Correlation
4x
10%
3x
15%
2x
20%
1x
25%
x Sep-03
Jan-04
May-04
Sep-04
Jan-05
Jun-05
30% Oct-05
Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
453
454 0.0%
0.2%
-4.6% -2.8% -1.6% -0.7%
3-7%
7-10%
10-15%
15-30%
0.5%
1.3%
2.3%
3.3%
0.6%
0.8%
-0.1%
-3.9%
Correlation Parallel -5.2%
-0.1%
15-30%
0.3% 0.4%
DJ CDX NA IG 10Yr (4/1/05-9/30/05) Average Average Curve Shift Tranche Spread Change 0-3% -2.2% -0.3%
-0.3%
10-15%
0.2% 0.1% 0.2%
-0.7%
7-10%
Correlation Parallel -2.8%
0.0%
-1.7%
3-7%
Average Curve Shift -0.5%
1.0%
1.7%
5.5%
-6.4%
Correlation Skew 0.0%
-0.9%
0.5%
1.9%
4.4%
Correlation Skew 0.0%
0.2%
0.2%
0.0%
0.0%
Spread Dispersion 0.1%
0.2%
0.0%
-0.1%
-0.3%
Spread Dispersion -0.8%
0.2%
0.7%
1.6%
2.2%
Time Decay 0.3%
0.2%
0.3%
0.4%
1.5%
Time Decay 3.8%
0.2%
0.4%
0.9%
2.4%
Carry 2.5%
0.0%
0.1%
0.3%
0.8%
Carry 2.5%
Tranche Performance Attribution Summary (Protection Seller Perspective)
2.0%
3.6%
7.3%
-7.0%
Total Return -4.8%
-0.4%
0.8%
2.3%
5.1%
Total Return -1.1%
chapter 65
DJ CDX NA IG 5Yr (4/1/05-9/30/05) Average Tranche Spread Change 0-3% -3.3%
exhibit 2
Structured Credit Insights – Instruments, Valuation and Strategies
Please see additional important disclosures at the end of this report.
455
-4.7% -1.8% -0.8% -0.3%
7-10%
10-15%
15-30%
Average Curve Shift -0.8%
DJ CDX NA IG 5Yr (Year to Date) Average Tranche Spread Change 0-3% -10.4%
3-7%
0.3%
1.3%
25-35%
0.0%
0.0%
0.1%
0.2%
0.6%
4.4%
0.8%
3.4%
15-25%
Average Curve Shift 0.2%
10-15%
DJ CDX NA HY (4/14/05-9/30/05) Average Tranche Spread Change 0-10% -1.1%
exhibit 2 (cont.)
0.1%
0.2%
0.4%
0.2%
Correlation Parallel -2.5%
2.5%
1.0%
-7.1%
Correlation Parallel -4.5%
-0.9%
0.4%
1.8%
4.7%
Correlation Skew 0.0%
-2.1%
3.8%
-2.4%
Correlation Skew 0.0%
0.3%
0.2%
0.3%
0.1%
Spread Dispersion -1.6%
2.4%
0.5%
-2.2%
Spread Dispersion -4.2%
0.3%
0.4%
0.7%
2.1%
Time Decay 5.2%
1.1%
1.6%
4.3%
Time Decay 2.1%
0.0%
0.1%
0.4%
1.1%
Carry 3.5%
0.7%
3.6%
0.0%
Carry 0.0%
Tranche Performance Attribution Summary (Protection Seller Perspective)
-0.5%
0.5%
1.8%
3.8%
Total Return -6.6%
6.2%
15.5%
-3.2%
Total Return -7.5%
456 -3.4% -1.7%
25-35%
Source: Morgan Stanley
-9.2%
DJ CDX NA HY (Year to Date) Average Tranche Spread Change 0-10% -9.5%
15-25%
Average Curve Shift 0.3%
-1.5%
15-30%
10-15%
0.5%
-3.5%
10-15%
0.4%
0.7%
1.0%
1.3%
2.2%
-6.5%
7-10%
3.0%
-11.0%
3-7%
Average Curve Shift -1.0%
2.1%
0.9%
-6.4%
Correlation Parallel -3.8%
0.8%
1.2%
0.0%
-5.8%
Correlation Parallel -8.3%
-0.1%
1.4%
-8.6%
Correlation Skew 0.0%
0.5%
2.0%
4.4%
-2.4%
Correlation Skew 0.0%
2.9%
1.1%
-2.2%
Spread Dispersion -6.9%
0.5%
0.6%
0.5%
0.0%
Spread Dispersion -0.2%
1.7%
3.1%
7.7%
Time Decay 4.2%
0.3%
1.0%
2.1%
2.9%
Time Decay 0.6%
1.0%
4.5%
0.0%
Carry 0.0%
0.2%
0.6%
1.3%
3.5%
Carry 3.7%
Tranche Performance Attribution Summary (Protection Seller Perspective)
6.2%
8.3%
-17.8%
Total Return -15.7%
1.4%
3.2%
4.1%
-9.8%
Total Return -11.7%
chapter 65
DJ CDX NA IG 10Yr (Year to Date) Average Tranche Spread Change 0-3% -6.4%
exhibit 2 (cont.)
Structured Credit Insights – Instruments, Valuation and Strategies
Correlation continued to be a key positive driver of investment grade mezzanine tranche outperformance, with the marked exception of the 10-year 3-7% tranche, which is behaving more like equity these days and hence has traded off. The correlation was enough to shift the absolute performance into negative territory just as it did with equity tranches. We argue that the divergent paths of equity and mezzanine returns are consistent with the changes in the underlying spread distribution for the CDX 4 index. In Exhibit 1, we show the now-familiar chart of the average spread for the 10 widest names divided by the average spread of the index. As we can see, the tail of the index increased markedly at the beginning of this year through April. This fattened tail effectively increases risk in equity tranches while reducing risk in mezzanine tranches because of the more concentrated default risk. It is also interesting that this tail tightened through the summer while equity correlation rallied; more recently, the tail widened again while implied correlation returned to the levels we experienced in May, both basically making a full round trip. The correlation impact for the entire period ( 5Y 0-3%) was -2.8% with a total return of -1.1%. However, the compounding moves in underlying spread dispersion and implied correlation made for volatile equity tranche returns within the six-month period. Mezzanine tranches, on the other hand, had spread effects basically muted by offsetting correlation moves throughout the period, making for more stable pricing. HIGH YIELD FEELS LESS TECHNICAL
Despite the increased activity in high yield tranches and several recent high yield default events, high yield tranches have not been immune to the technical environment that is driving performance in the IG tranched space (see Exhibit 2). Spread changes, a default event, roll down and correlation were all drivers of returns. It is worth noting that the impact of the Collins & Aikman default (captured in the dispersion impact) was largely offset by other factors for all but the 0-10% tranche, which actually had to make a loss payment. We would attribute this to the higher absolute default risk priced into high yield spreads (see Chapter 38). Pricing for most investment grade tranches other than first loss is largely an exercise of allocating the risk premium in the underlying market, because expected losses in all but the first loss tranches are rather low. A much larger proportion of the spread in mezzanine high yield tranches can be associated with expected losses.
Please see additional important disclosures at the end of this report.
457
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 65
exhibit 3
Historical Expected Loss/Market Implied Loss for Various Tranches
120% 5Y HY 14.6524.65%
5Y HY 9.65-14.65%
100%
80% 10Y IG 0-3%
60% 7Y IG 0-3%
40%
5Y IG 0-3% 5Y HY 24.6534.65% 7Y IG 7-10%
20%
5Y IG 3-7% 7Y IG 3-7% 10Y IG 3-7% 10Y IG 7-10%
0% 0
10Y 10-15%
500
1,000
1,500
2,000
2,500
Spread (bp)
Source: Morgan Stanley
To illustrate the point, in Exhibit 3 we summarize the ratio of expected losses as estimated by our historical default rate approach (see Chapter 32) to aggregate losses implied from tranche spreads for the equity and mezzanine tranches. The higher this ratio, the greater the portion of the total risk that can be thought of as realizable default risk (based on our limited 30-year history) versus the additional risk premium investors charge for taking that risk. We see that the ratio is generally much higher for the high yield tranches when compared to investment grade tranches with similar spreads. The market implications of these higher ratios is that a bigger part of the all-in price for high yield mezzanine is associated with possible losses (presumably because of volatile historical default experience), especially when compared to investment grade tranches, for which the ratios are much lower. CONCLUSION
The recent divergence in performance between equity and mezzanine tranches in both investment grade and high yield tranches highlights how changes in implied correlation can compound and/or mute the impact of changes in underlying spreads across the capital structure. Clearly, investors remain very comfortable with mezzanine risk, while equity pricing has reflected the more idiosyncratic nature of recent market events. DELPHI UPDATE
Following the bankruptcy filing of Delphi Corp. on October 19, S&P downgraded 126 synthetic CDO classes and placed an additional 39 classes on negative watch. It is estimated that just under 800 public and confidentially rated European and North American synthetic CDOs referenced Delphi at the time of the credit event. The average downgrade severity for all of the classes (European and North American deals) was 1.7 notches, with 52% of the tranches receiving a single-notch downgrade and 33% receiving a two-notch downgrade. Fifteen percent of the classes was downgraded three to five notches.
458
For both the United States and Europe, the 2004 vintage accounted for the largest portion of downgrades. Forty-six percent of European synthetic CDO downgrades and 45% of North American downgrades were from the 2004 cohort. Thirty-one percent of the European and 29% of the North American downgrades were from the 2005 vintage. S&P noted that Delphi was not the only factor that accounted for these downgrades. Though it played the most significant role, other corporate downgrades, such as the Neiman Marcus Group, Ford Motor Co. and General Motors Corp,. also impacted the credit quality of the underlying portfolios.
Please see additional important disclosures at the end of this report.
459
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 66
Modifications to S&P Synthetic CDO Ratings Model
December 20, 2005
Primary Analyst: Vishwanath Tirupattur Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Pinar Onur Standard & Poor’s has launched a new version of its CDO modeling tool for synthetic CDOs – CDO Evaluator (Version 3.0). The new version will be used in conjunction with the current version (Version 2.4.3) during a transition period that ends on March 31, 2006, after which the new version will become the sole version. According to S&P, the motivation behind these changes is to bring the assumptions embedded in the Evaluator model in line with new historical transition and default data. S&P also noted that the new version of the model will not be used for cash CDO transactions until early 2006, when a separate transition process will be announced for them. As a consequence of the introduction of the new version of the model, S&P placed 35 tranches from 18 US synthetic CDO transactions on CreditWatch with negative implications and 15 tranches from nine European synthetic CDO transactions on CreditWatch with positive implications. Based on publicly available information on these transactions, it seems that transactions with predominantly high yield corporate/emerging market collateral have been put on negative watch and those with predominantly investment grade underlying collateral have been placed on positive watch. S&P noted that it expects to resolve the CreditWatch placements within 90 days. Barring any restructuring that may occur during this period, S&P observed that approximately half of the potential downgrades would be one-notch downgrades and the rest would be evenly split between two- and three-notch downgrades. WHAT HAS CHANGED?
New Default Tables: Default probabilities for corporates, sovereigns, CDO securities and asset-backed securities have changed. We demonstrate the differences in default probabilities across various rating categories for corporates at the 7-year point in Exhibit 1. For corporates, default probabilities in the new version of the model are lower for all investment grade ratings categories except BBB- and higher for all below investment grade categories except BB+ and BB categories. Default probabilities for CC and below are same (100%) under both versions of the model. Term Structure Effects: There are significant differences in the term structures of default probabilities between the two models. For corporates, for higher ratings categories within investment grade, the term structure of default probabilities is significantly flatter in absolute terms in the new version, suggesting a smaller marginal increase in absolute level of default probabilities with an increase in maturity. For example, default probabilities for “A” rated corporates in the old version increased from 0.136% to 3.041% as maturity increased from 1 year to 10 years. In the new version, the same default probability increases from 0.022% to 1.782% (Exhibit 2).
460
exhibit 1
7-Year Corporate Probability of Default
8% Old
7%
New
6% 5% 4% 3% 2% 1% 0% AAA
AA+
AA
AA-
A+
A
A-
BBB+
BBB
BBB-
CCC
CCC-
CC
Obligor Rating 100% 90%
Old
80%
New
70% 60% 50% 40% 30% 20% 10% 0% BB+
BB
BB-
B+
B BCCC+ Obligor Rating
Source: S&P and Morgan Stanley calculations
On the other hand, on a relative basis, a different picture emerges. From the same numbers above, the default probabilities increased by a factor of 22 for “A” rated corporates in the old version of the model as the maturity term increases from 1 year to 10 years. In the new version, they increase by a factor of 81. As such, the effect of maturity extension on CDO tranche ratings for transactions with investment grade collateral is not entirely clear at this stage and warrants further analysis. However, for certain below investment grade categories, the default probability curves become steeper with time. For “BB” rated corporates, the default probabilities for maturity terms 8 years and above are higher in the new version. For “B” rated corporates, the term structure is steeper in the new version from the 3-year maturity point and is getting steeper with increasing maturity. This suggests that maturity extension for lower credit quality corporate portfolios could result in higher attachment points. All things equal, existing transactions with high yield collateral with a long remaining term to maturity (for example, 7–10 years) are likely to be affected negatively, while the effects on transactions with investment grade collateral are likely to be more constructive.
Please see additional important disclosures at the end of this report.
461
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 66
exhibit 2
3.5%
Term Structure Effects
3.0%
Old
2.5%
New
Default Probabilities for "A" Rated Corporates
2.0% 1.5% 1.0% 0.5% 0.0% 1
2
3
4
5
6
7
8
9
10
Years 45%
Default Probabilities for "B" Rated Corporates
40% 35%
Old
30%
New
25% 20% 15% 10% 5% 0% 1
2
3
4
5
6
7
8
9
10
Years
Source: S&P and Morgan Stanley calculations
ABS Securities: For ABS securities, the new version incorporates maturity dependent default probability tables in place of a common set of default probabilities for all maturities under the current version. In Exhibit 3, we plot the difference between default probabilities for various ratings categories of ABS collateral at three, five and seven year maturity points. Default probabilities are in general lower for all investment grade ratings categories until BBB for 3-, 5- and 7-year terms in the new version. For tranches BB- and below, default probabilities are higher for 5- and 7-years and for all terms for categories CCC+ and below. Sovereigns and CDO Securities: As was the case with the previous version, the new version will use corporate default probability tables for sovereigns. Since corporate tables have changed in the new version, sovereign tables also have changed accordingly. In the previous version, default probabilities were the same for CDO securities and corporates. In the new version, a separate table has been created for CDO securities with default probabilities that are higher than corporate default probabilities across ratings categories and maturities. Correlation: For corporates, inter-sector correlations have increased to 5% from 0% for sectors within a country and within a region. Inter-sector correlations between regions remain at 0%. Intra-sector correlations have decreased from 30% to 15% within a country and from 15% to 0% within a region with local classification. The remaining correlation assumptions are largely unaffected.
462
The consequences of the changes in correlation assumptions are somewhat ambiguous for investment grade synthetic CDOs and the ratings impact on CDO tranches is a function of the rating and maturity of the underlying collateral. Lower default probabilities may be partially offset by higher correlations for investment grade collateral. For transactions with high yield collateral, the increase in correlations will compound the negative consequences of higher default probabilities and might result in higher attachment points. Finally, the adjustment factors for making rating-dependent adjustments to portfolio default/loss rates have been removed. CONCLUSION
The changes in default probabilities and correlations are meaningful. By and large, the changes are constructive for investment grade synthetic CDOs and negative for high yield synthetic CDOs. However, there are significant term structure effects that need to be carefully considered. We expect that at least some of the transactions affected by these changes will be restructured. For ratings sensitive investors, it might be useful to revisit their ratings analysis under the new version and consider such restructuring opportunities. exhibit 3
ABS Probability of Default (New-Old)
2.0% 3 Yr
1.5%
5 Yr
7 Yr
1.0% 0.5% 0.0% -0.5% -1.0% -1.5% -2.0% AAA
AA+
AA
AA-
A+ A Obligor Rating
A-
BBB+
BBB
BBB-
CCC+ CCC
CCC-
CC
30% 25%
3 Yr
5 Yr
7 Yr
20% 15% 10% 5% 0% -5% -10% BB+
BB
BB-
B+
B
B-
Obligor Rating
Source: S&P and Morgan Stanley calculations
Please see additional important disclosures at the end of this report.
463
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 67
Rating Between the Lines
January 20, 2006
Primary Analyst: Sivan Mahadevan Primary Analyst: Vishwanath Tirupattur Primary Analyst: Peter Polanskyj Primary Analyst: Pinar Onur Whether or not one focuses on ratings, actions taken by agencies have profound impacts on the credit markets today. The recent focus started when the fixed income industry’s most popular index provider decided to include Fitch ratings in its criteria last year (see our investment grade strategy team’s report “Change of Rules, Chain of Fools,” January 28, 2005). While this shift slowed the process of moving the US auto manufacturers out of popular corporate bond indices, eventually all of the agencies had no choice but to downgrade the parent company and credit subsidiaries, and the market had to digest the event that everyone feared, namely the handoff of significant amounts of auto credit risk from investment grade to high yield investors. Ironically, a potential GMAC sale to an investment grade rated institution could reverse a decent amount of this flow. Within the structured credit world, S&P introduced a new version of its popular CDO ratings model in late December (CDO Evaluator 3.0), the ramifications of which will be felt in both investment grade and high yield markets in the coming months (see Chapter 66. Effectively, S&P has adjusted the assumptions that it makes for default probabilities along both the ratings and maturity spectrums, in line with its updated data on historical ratings transitions and defaults. In general, investment grade default rate assumptions fall (except for BBB-), while high yield assumptions rise (except for BB and BB+, for which default probabilities in the new model are slightly lower). In the high yield space, the model impacts are significant. While the bespoke high yield structured credit activity in the unsecured space is limited, there are some implications for index tranches as well as underlying credits. We review our thoughts on them. In the investment grade space, the model impacts are less severe, but the market impact is potentially huge, and the deal pipeline is already beginning to reflect this. We go through the details in this chapter and opine on the impact. On the positive side, there is justification for the actions S&P took, and future deals may actually be healthier creditwise as a result. But one still has to be careful with leverage, as it just got a bit easier to use.
464
Moody’s, on the other hand, has looked elsewhere. After having taken the large step last year of moving from its 10-year-old binomial-expansion model to modern day correlation models, it is now focused on the vastly different recovery outcomes for credits in the market place. As the Moody’s approach is based on expected losses (whereas S&P focuses only on probability of default), Moody’s is reaching out to the market today to solicit feedback on a methodology that could push ratings higher on senior secured high yield loans compared to unsecured high yield bonds, which in turn can have important ramifications for the ratings of outstanding CLOs and new CLO structures (see “Probability-of-Default and Loss-Given-Default Ratings for Non-Financial SpeculativeGrade Corporate Obligors,” Moody’s Investors Service, January 2006). exhibit 1
5 Year Default Rates – S&P Old and New
6% S&P (New CDO Model) S&P (Old CDO Model)
5%
4%
3%
2%
1%
0% AAA
AA+
AA
AA-
A+
A
A-
BBB+
BBB
BBB-
Source: Morgan Stanley, Moody’s, S&P
INVESTMENT GRADE – IS A MORE CONSTRUCTIVE DEFAULT VIEW JUSTIFIED?
Nearly five years have passed since the peak of the fallen-angel default activity in the last recession, but still no day goes by without us thinking about how such default activity can impact structures that we see in the market today. Yet, we have said repeatedly that investment grade credit has exhibited positive risk premiums through most parts of the credit cycle (including 2001–02), meaning that actual default rates on average are lower than those implied by market spreads, even Libor spreads. This popular belief has kept underlying spreads tight, in part because structured credit vehicles have been used to implement the popular view. We have not met an investor yet who feels that a 7% attachment point on an investment grade CDX-like portfolio could be touched with losses from defaults over a 5-year period. This in turn justifies 7–10% type tranche pricing moving from above 300 bp in 2003 to below 30 bp today. But sampling risk is still key (see our thoughts at the end of this chapter).
Please see additional important disclosures at the end of this report.
465
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 67
INVESTMENT GRADE COLLATERAL
The S&P approach is motivating a new type of synthetic CDO in the market. Leverage on higher quality names will be encouraged, while the inclusion of credits that show any sign of falling to BBB- (let alone being rated BBB-) will likely be reduced to a minimum. This latter phenomenon is probably a good thing, as the likelihood of credits falling below investment grade is significantly reduced. Using historical ratings transition data from Moody’s (we don’t have it for S&P), the probability of a Baa2 rated credit falling to high yield (including default) is about 11% over 5 years, while for a Baa3 credit it is 22%, double the rate (see Exhibit 2). The 5-year default rate difference is also double (1.5% vs. 3%). exhibit 2
Historical 5-Year Transition Rate to High Yield
25%
20%
15%
10%
5%
0% Aaa
Aa1
Aa2
Aa3
A1
A2
A3
Baa1
Baa2
Baa3
Source: Morgan Stanley, Moody’s
Interestingly, while the impact of including BBB- names in portfolios is clearly negative relative to the prior model, including high quality high yield is actually encouraged (or neutral) in the new S&P model. One reasonable interpretation is that BBB- credits are on their way down and BB or BB+ credits are on their way up in credit quality terms, which is consistent with the views expressed by our strategy team in the past. MODELING CHANGES – CORRELATION
On the structural front, S&P has reduced correlation for credits within sectors, increased the correlation of sectors with one another, and reduced some correlation assumptions both within and between geographic regions. As such, relative to older structures, we expect new CDOs to have the following characteristics:
466
•
More credits within a given sector,
•
Fewer sectors, and
•
More global portfolios.
Whether the implication of these changes is positive or negative depends on one’s perspective. Under the old model, higher ratings were achieved in part by reaching for diversification at the sector level. We will likely see less of that now, which means that portfolios could begin to resemble debt market capitalization weighted corporate bond indices, given the disproportionate weighting of sectors. However, higher intra-sector concentration is not necessarily good for all rated tranches. On the other hand, more global portfolios should add the kind of “real” diversification that is valuable to rated tranches. THE NEW INVESTMENT GRADE STRUCTURES
Lower default probabilities for most investment grade credits do not directly translate into lower attachment points at the CDO structure level, as the lower default probabilities apply to the liability side as well. In fact, if we consider the CDX 5 portfolio (which has 12 BBB- rated credits out of 125), we find that the average attachment point increases to maintain the same rating in the new model (roughly 0.3%, see Exhibit 3). However, the more important impact will likely be in bespoke structures for which the credit and sector composition can be optimized for the new ratings regimes and can thus generate lower attachment points. Higher in the capital structure, the model changes have less impact, at least for the CDX portfolio. In some respects, we’ve already seen this occurring with index tranches for CDX Series 4 and 5, which trade flat on an absolute spread basis despite notable differences in portfolio spreads and different maturities. exhibit 3 Rating Bucket AAA
IG CDX 5 – Ratings-Based Attachment Points Would Rise, But Could Fall for Other Portfolios Average of 5 Year 0.1%
Average of 10 Year 0.1%
AA
0.4%
0.4%
A
0.5%
0.6%
BBB
0.1%
0.4%
BB
0.1%
0.5%
B
0.3%
0.2%
Average
0.3%
0.4%
Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
467
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 67
HIGH YIELD COLLATERAL
There is a small market for purely synthetic high yield CDOs, and in our view there is a lot more outstanding risk in the index tranches than in bespoke. S&P ratings changes will have a net negative impact on ratings for HY synthetic CDOs which will result in two things. First, the bespoke HY synthetic CDO market will remain small (especially with a newfound focus on investment grade CDOs and the potential for a synthetic CLO market to develop), and second there may be some interesting pricing implications on the HY index tranche market and more generally high yield credits. We do note, however, that BBs (BB+ and BB) could still appear in bespoke structures with majority investment grade collateral, given the slight decrease in their default probabilities in the new model. For credits rated BB- or lower, their inclusion in synthetic CDO portfolios will become costlier in ratings terms. To the extent that the structured credit bid was contributing to keeping these spreads tight, the deterrence introduced by the new S&P model is negative for them. Focusing on the index tranches, we estimate that for a HY CDX 5 portfolio, a BBBattachment point rises from about 15% to 20%, while a BBB attachment rises from nearly 17% to 22% (see Exhibit 4). Today’s 15–25% tranche (which is really a 13.6% attachment following the two defaults) is really not an investment grade rated tranche, which may explain why it has continued to rally (driven by time decay) despite the changes to the S&P model. Market participants have also argued that this rally is based on bespoke hedging activity in the 20–25% sub-tranche. With the average impact of the model change indicating about 5% higher attachments broadly, this argues for less demand to hedge risk from bespoke origination in the 15–25% benchmark tranche Therefore, we would argue that 15–25% is likely to begin trading more like an OTM option on HY defaults (in today’s environment) than based on ratings. The ratings impact higher up in the capital structure is no less important with 25–35% tranches moving from AA-type ratings to BBB-type ratings. The market may have been ahead of S&P on the pricing of these tranches, as they have traded quite wide when compared to AA’s with different underlying collateral. That being said, the senior HY tranches are just as likely to be affected by pricing on the equity tranches and arbitrage forces as ratings. The large spread differential between series 4 and series 5 25–35% tranche spreads supports this and provides a counter example to what we see in IG 5 year senior tranches. exhibit 4 Rating Bucket AAA
Average of 5 Year 3.2%
Average of 10 Year 3.7%
AA
4.5%
5.4%
A
5.4%
6.7%
BBB
5.1%
7.1%
BB
5.7%
8.0%
B
5.4%
7.3%
Average
4.9%
6.4%
Source: Morgan Stanley
468
HY CDX 5 – Ratings-Based Attachment Points Would Rise Significantly
CONCLUSION – RATINGS DIFFERENCES, SAMPLING RISK AND SPONSORSHIP
The approaches the agencies adopt in the structured credit world have important market implications, and by the same token, the continued growth of the structured credit businesses is important to the agencies (as discussed in the Wall Street Journal on January 19, 2006), so they are likely to carry on with new development in this space. There will continue to be ratings differences between S&P and Moody’s in the synthetic CDO space, given their different philosophies (probability of default versus expected loss) and of course modeling differences and subjective issues. For investors requiring both ratings, the frustration does not go away, but in the grand scheme of things it doesn’t hurt to have two distinct opinions about credit quality. We also highlight that rating agency models continue to treat credits with the same ratings similarly, which creates more homogeneity than we see in the market. As such, portfolio sampling risk, a topic to which we have devoted a lot of research, will always be a key risk in any given transaction (see Chapter 32). For the underlying credit markets, we expect S&P modeling changes to result in much less sponsorship for weak BBBs, unless of course their spreads are attractive enough to warrant the ratings hit. The same is true for B-rated issuers, with BBs being the potential beneficiary.
Please see additional important disclosures at the end of this report.
469
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 68
CDO Equity Performance – Unplugged Primary Analyst: Vishwanath Tirupattur Primary Analyst: Sivan Mahadevan Andrew Sheets As structured credit products researchers, we are frequently asked to characterize the realized returns performance of CDO equity as an asset class, as well as how performance has varied across different types of CDOs and vintages. Typical CDO pitches include proclamations pertaining to the timing of CDO equity returns and the consistency of returns of some types of CDOs relative to others, and so on. We are often asked if they have indeed been realized over time. For an asset class that has been around for over 10 years, with record issuance volumes in the last two years, it is somewhat surprising that answering questions about such return characteristics has not been easy thus far. The main constraint has been the sheer absence of data on equity distributions in the public domain. Without actual data on returns, the approaches pursued have included various hypothetical analyses that provide only limited value in addressing such questions. Some studies have used actual return distributions, but those were based on relatively small sample sizes, which makes it difficult to draw more generalized conclusions. In late March 2006, Moody’s partially lifted the veil of ambiguity from CDO returns by making equity distributions data available on 603 transactions. The data are at the transaction level and include total cash distributions paid to equity tranches. While the identity of transactions is not public, vintage year and CDO type characterize each. Although further data on the timing of distribution and identity of deals would be useful, we still consider the data provided useful in attempting to meaningfully characterize CDO equity returns. In our judgment, this is a positive development towards enhancing the transparency of the CDO market. Analysis of Moody’s data on equity distributions is the focus of this chapter. COMPUTING EQUITY RETURNS: NOT AS EASY AS IT SEEMS
If you have data on the original purchase price, equity distribution amount, dates of distribution and their terminal value, it is straightforward to compute realized IRRs – but there’s the rub. Even with the data that Moody’s has unveiled, we only have information on total distributions to equity tranches and not on any of the rest of the elements. As most market participants are aware, the original purchase price is a privately negotiated level. While, assuming the original purchase to be at par seems reasonable, we know that discounts are not uncommon at deal inception. Limited secondary market activity on equity tranches also means there is no publicly discernible data on terminal values of these tranches. Obviously for terminated transactions we can safely assume that data on distributions includes payments made, if any, at deal termination. As such, we do not have all the necessary data for calculating precise realized returns. We will go along with the Moody’s approach of assuming the original price to be par and computing returns using Microsoft Excel’s XIRR function for terminated transactions. However, we think that the absence of terminal prices for non-terminated transactions is significant enough that we cannot use Moody’s computed XIRRs but
470
must instead focus on total cash on cash returns. Since we also do not have distribution dates, we use undiscounted values as reported by Moody’s. SUMMARY OF RESULTS
•
Realized IRRs for terminated CDOs have been disappointing, especially in HY CBOs (median IRR of -19.5%), where only two out of 14 deals recovered their initial principal. Median IRRs for terminated deals were positive but still quite low for EM CBOs (2.4%) and CLOs (1.8%).
•
While it may be surprising that realized IRRs were less than stellar for terminated CLOs, the potential for refinancing due to the historically low funding costs for new issue CLOs may be the motivation behind their termination.
•
Equity returns from terminated CLOs outperformed terminated HY CBO equity over a similar period, posting both higher average returns and less performance volatility. Still, only a little over half of terminated CLO equity deals had positive IRRs.
•
In non-terminated deals, CLOs consistently outperformed CBOs, with higher average returns and significantly more deals recovering their initial investment. SF CDO deals have less historical data, but it appears that early vintage deals struggled while the performance of more recent transactions (2003-2004) has been strong.
•
Equity distributions, particularly for CBOs, were front-loaded, with a significant part of their return realized in the first three years. CBOs experienced a significant drop-off in returns after the deal’s third year, while a gradual decline in distributions was experienced by CLOs and SF CDOs. Both CLOs and SF CDOs showed less variability in returns with the passage of time.
•
Annual distributions across CLOs and SF CDOs were generally dispersed, regardless of vintage, suggesting wide differentiation between the best and worst performers. This indicates that security selection, and therefore manager selection, really matters. The distribution of CBO returns was, by contrast, highly skewed to the negative.
Moody’s data includes 66 transactions that have fully terminated. In Exhibit 1, we provide descriptive statistics of their realized IRRs, aggregated by CDO type. 1 exhibit 1
Realized IRRs of Terminated CDOs HY CBO 14
EM CBO 11
CLO 32
IG CBO 2
SF CDO 7
Mean
-19.5%
1.5%
-2.6%
11.1%
2.7%
Median
-17.3%
2.4%
1.8%
11.1%
5.7%
Maximum
98.5%
22.9%
28.0%
22.6%
45.0%
Minimum
-81.7%
-40.6%
-59.0%
-0.4%
-51.7%
Standard Deviation
47.7%
18.7%
21.6%
16.2%
33.4%
2
7
18
1
5
No. of deals
# of Deals with positive IRRs
Source: Moody’s; Morgan Stanley Calculations
1
We excluded data on the two investment-grade CBOs and seven structured-finance CDOs.
Please see additional important disclosures at the end of this report.
471
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 68
At an aggregate level, the realized IRRs are disappointing but not entirely surprising, given that the terminated transactions were from vintages that experienced a gutwrenching credit cycle with the worst default rates in nearly 70 years. Median returns were slightly positive for EM CBOs and CLOs and significantly negative for HY CBOs (ignoring IG CBOs and SF CDOs due to sample size considerations). Only two of the 14 HY CBOs had positive realized IRRs, while 18 of the 32 CLOs and seven out of 11 EM CBOs had positive IRRs. Given the continued popularity of CLOs, it is a little surprising that the median equity returns on terminated deals were far from stellar. We believe a big driver of this phenomenon is the motivation for reinvestment into newer transactions where funding costs (i.e., mezzanine and senior tranche spreads) are significantly lower. For investors with a constructive view on CLO equity, reinvesting in a new structure remains worthwhile, even if that means realizing a less than optimal return on the original transaction. More interesting is the distribution of returns across deals and the differences therein across CDO types. In Exhibit 2, we compiled the realized IRRs of terminated HY CBO and CLOs. What we find striking about this distribution is how different the experiences of equity holders in CLO and CBOs have been. Equity returns for CBOs were not only disappointing, but also highly volatile, with IRRs broadly distributed in a range from 81.7% to +98.5%. CLO equity, at least by comparison, was a far superior asset class, with higher average IRRs obtained with markedly less volatility. We find these results from called deals consistent with our analysis of the broader universe of data, where CLO equity posted returns with lower average volatility at each year of seasoning. exhibit 2
Distribution of Realized IRRs: CLOs vs. CBOs
% of Deals 50%
CBOs CLOs
45% 40% 35% 30% 25% 20% 15% 10% 5%
10 0%
80 % ->
->
->
-> 80 %
60 % 60 %
40 % 40 %
20 % 20 %
0% ->
-> 0%
-2 0%
-2 0%
-4 0%
-> -4 0%
-> -6 0%
-> -8 0%
-1 00 %
->
-8 0%
-6 0%
0%
Source: Moody’s; Morgan Stanley Calculations
Note that relative to measures of central tendency of the realized IRRs (mean, median, etc.), their measures of central dispersion (standard deviation, min-max range) were significantly larger. If we knew the identities of these transactions, we could attribute this variability to manager or structure or vintage. Unfortunately, the available data limit our ability to tell a fuller story.
472
NON-TERMINATED TRANSACTIONS
We analyze the cash on cash returns for the non-terminated transactions. In Exhibit 3, we summarize the undiscounted total distributions received by equity tranches by CDO type and their vintage years. We illustrate the data in Exhibit 3 with an example. Moving down the column for the vintage year 2000, the data covers 17 CLO equity tranches, which, on average, received total distributions equal to 92% of the face amount of the equity tranche. Of the 17 CLO equity tranches, the highest amount received was 157% and the lowest amount received was 34%, both expressed as a percentage of the face amount of the tranches. Forty-seven percent of these 17 CLOs received distributions totaling amounts higher than 100% of their face amount – or got their original investment back (assuming the tranches were priced at par at deal inception). exhibit 3
Total Distributions to Equity Tranches by CDO Type and Vintage Year Vintage Year 1996 1997 1998 1999 2000 2001 2002 2003 2004
CLO No. of deals Mean Median Maximum Minimum Std Dev % of Deals with Distributions > 100% SF CDO No. of deals Mean Median Maximum Minimum Std Dev % of Deals with Distributions > 100% HY CBO No. of deals Mean Median Maximum Minimum Std Dev % of Deals with Distributions > 100% IG CBO No. of deals Mean Median Maximum Minimum Std Dev % of Deals with Distributions > 100%
2 3 12 16 17 15 29 103% 82% 59% 84% 92% 86% 75% 103% 48% 56% 76% 90% 93% 69% 149% 156% 145% 159% 157% 127% 291% 57% 41% 9% 13% 34% 29% 12% 65% 64% 36% 37% 36% 28% 49% 50% 33% 17% 31% 47% 33% 10% 2 19 27 31% 55% 51% 31% 44% 49% 37% 157% 103% 26% 13% 8% 8% 40% 25% 0% 11% 4%
37 43% 46% 85% 1% 23% 0%
34 32% 30% 77% 10% 13% 0%
41 15% 15% 29% 3% 6% 0%
39 38% 38% 69% 5% 17% 0%
54 19% 16% 47% 3% 11% 0%
5 11 24 45 30 30 4 118% 46% 40% 38% 42% 73% 78% 84% 53% 42% 29% 32% 80% 83% 211% 119% 139% 205% 110% 134% 106% 56% 7% 8% 0% 8% 11% 39% 66% 32% 29% 34% 28% 35% 29% 40% 9% 4% 4% 3% 20% 25% 2 3 10 135% 84% 39% 135% 44% 24% 177% 171% 180% 92% 39% 14% 60% 75% 50% 50% 33% 10%
11 4 54% 70% 50% 64% 86% 107% 28% 44% 20% 30% 0% 25%
Source: Moody’s; Morgan Stanley Calculations
Please see additional important disclosures at the end of this report.
473
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 68
For nearly all vintages with a reasonable number of transactions, CLOs outperformed CBOs by a significant margin. Also with CLOs, with the exception of 1998, for all vintages with more than five years of existence, about one-third of transactions have already received their original investments. Comparable numbers are depressingly low for HY CBOs. With vintages 1997-2000, more than 90% of the HY CBO transactions have not yet received their original investment. The likelihood of any future distributions for most of these transactions is slim. The contrasting performance of CLOs versus CBOs is telling. Even in a terrible vintage year like 1998, the worst performing CLO vintage year, average distribution to CLOs was 13% higher for CLOs compared to HY CBOs. SF CDOs have a relatively shorter history since their issuance did not start in earnest until the year 2000. SF CDOs of vintages 2000-2002 performed quite poorly, with nearly 90% of them yet to receive their full face value. Their subsequent vintages show a significant pick-up in equity distributions. The performance of SF CDOs, dismal for vintages 2002 and prior and a significantly improving story for subsequent vintages, is directly related to their collateral. See Chapter 54 for a detailed exposition of this topic. For the relatively recent vintages of 2003 and 2004, SF CDOs outperformed CLOs, based on aggregate average distributions. It is noteworthy that the variability across CLOs was significantly lower than the variability across SF CDOs. Note the differences in the standard deviations of their distributions – much lower for CLOs. We posit the relative uniformity of credit bets across CLO portfolios as a plausible explanation for the limited variation across CLOs (see Chapter 56 for fuller analysis). DEAL AGE ANALYSIS
Deal age analysis is an examination of cash flow by transaction age across vintages. Exhibit 4 shows the analysis aggregated by three important CDO types. For example, across all vintages years for which data are available, the average first year distributions were 13% for CBOs and SF CDOs and 12% for CLOs. exhibit 4 Deal Age (Yrs) 1
CBOs Average 13%
Deal Age Analysis
St Dev 3%
St Dev 3%
SF CDOs Average St Dev 13% 4%
2
21%
11%
19%
6%
15%
6%
3
21%
24%
17%
6%
13%
5%
4
9%
8%
14%
7%
9%
7%
5
4%
4%
10%
9%
13%
13%
6
2%
2%
8%
8%
15%
17%
Source: Moody’s; Morgan Stanley Calculations
474
CLOs Average 12%
Based on these numbers, there seems to be some support for the notion that equity distributions are front-loaded for CDOs. Exhibit 5 demonstrates this effect pictorially. The support for this notion is strongest for CBOs where distributions drop off sharply after year 3. Even though distributions to CLO equity tranches also decline after year 3, the decline is not as steep. In fact, the relative stability of CLO distributions across deal age is worth noting. We caution against the implication here that distributions for SF CDO equity tranches increased on average in years 5 and 6, given that the data for SF CDOs aged 5 or 6 years is based on a small sample of observations. Note that the variability of equity distributions across deal age is significantly lower for CLOs and SF CDOs relative to HY CBOs. For example, distributions in year 3 for CBOs average 21%, versus 17% for CLOs and 13% for SF CDOs. However, the variation among CBOs with an attained age of three years is huge (24%) relative to CLOs (6%) and SF CDOs (5%). exhibit 5
CDO Equity Distributions 1996-2004
% of Notional Returned 25%
CBOs CLOs SF CDOs
20%
15%
10%
5%
0% 1
2
3 4 Deal Age (Yrs)
5
6
Source: Moody’s; Morgan Stanley Calculations
MANAGER MATTERS
In Exhibits 6-9 on the following pages, we depict frequency distributions of the average annual equity distributions 2 of different CDO types by vintage. For example, in Exhibit 6, 25% of the 1996-98 vintage CBOs had average annual distributions in a 0-2.5% range. In the case of HY CBOs (Exhibit 6), the distributions are skewed strongly to the left (low single digit returns) for vintages 1996-98 and 1999-00. However, for 2001-02, distributions are significantly more evenly distributed. It is ironic that the vintages that marked the end of their issuance saw some of the best-performing CBOs. The degree of skewness to the left is significantly more muted for CLOs (Exhibit 7), if we exclude 199698 vintage of transactions. Distributions are actually skewed to the right (high teens and higher) for 2001-02 vintages. For SF CDOs, the vintage effect is clear. The performance of 1999-00 vintages was significantly skewed to the left. The 2001-02 vintage showed similar performance, albeit to a lesser extent. 2003-04 vintages appear less skewed. Data for IG CBOs are relatively small, but the dispersion in the returns is noteworthy. 2
We compute annual equity distributions for each transaction by dividing the total distributions by the age of the transaction.
Please see additional important disclosures at the end of this report.
475
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 68
Exhibits 6-9 highlight that performance is widely dispersed across vintages and across CDO types. Considering that the structures have been relatively stable, particularly within vintage years, this dispersion suggests that the differences in performance are attributable to security selection and therefore to the manager. exhibit 6
HY CBO Annualized Cash Distributions by Vintage
% of Populations
1996-98
45%
1999-00
40%
2001-02
35% 30% 25% 20% 15% 10% 5%
02. 5 2. 55 57. 5 7. 51 10 0 -1 2. 12 5 .5 -1 15 5 -1 7. 17 5 .5 -2 20 0 -2 2. 22 5 .5 -2 25 5 -2 7. 27 5 .5 -3 0 30 -1 00
0%
Annual Return Bucket (%)
Source: Moody’s; Morgan Stanley Calculations
exhibit 7
CLO Annualized Cash Distributions by Vintage
% of Populations
1996-98
40%
1999-00
35%
2001-02 2003-04
30% 25% 20% 15% 10% 5%
02. 5 2. 55 57. 5 7. 51 10 0 -1 2. 12 5 .5 -1 15 5 -1 7. 17 5 .5 -2 20 0 -2 2. 22 5 .5 -2 25 5 -2 7. 27 5 .5 -3 0 30 -1 00
0%
Annual Return Bucket (%)
Source: Moody’s; Morgan Stanley Calculations
476
exhibit 8
IG CBO Annualized Cash Distributions by Vintage
% of Populations 2001-02
25% 20% 15% 10% 5%
02. 5 2. 55 57. 5 7. 510 10 -1 2. 5 12 .5 -1 5 15 -1 7. 5 17 .5 -2 0 20 -2 2. 5 22 .5 -2 5 25 -2 7. 5 27 .5 -3 0 30 -1 00
0%
Annual Return Bucket (%)
Source: Moody’s; Morgan Stanley Calculations
exhibit 9
SF CDO Annualized Cash Distributions by Vintage
% of Populations 25%
1999-00 2001-02 2003-04
20%
15%
10%
5%
02. 5 2. 55 57. 5 7. 510 10 -1 2. 5 12 .5 -1 5 15 -1 7. 5 17 .5 -2 0 20 -2 2. 5 22 .5 -2 5 25 -2 7. 5 27 .5 -3 0 30 -1 00
0%
Annual Return Bucket (%)
Source: Moody’s; Morgan Stanley Calculations
CONCLUSIONS
The equity distributions data released by Moody’s is a significant first step in enhancing transparency in cash CDO returns. There is a wide variation across CDO types and vintages in the performance of CDO equity tranches. As coverage expands and additional details are made public, the opportunity to enhance our understanding of CDO returns will increase. We are sure to return to this topic. We acknowledge the contributions of Simmi Sareen to this chapter.
Please see additional important disclosures at the end of this report.
477
Section F
Cash and Synthetic Convergence Themes
Please see additional important disclosures at the end of this report.
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 69
Bridging the High Yield Gap
September 5, 2003
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Anisha Ambardar Most of the recent progress in correlation trading has been focused on investment grade credit. Interest and growth here seems natural, given the rapid development of the underlying derivatives market over the past few years. Yet, with the increasingly evident shift in the US economic cycle, we argue that structured high yield credit may be poised for some increase in popularity going forward. While we can point to only a limited amount of investor interest to date, some of the synthetic infrastructure has been built, thanks to a benchmark high yield basket product and existing correlation models. So, if investors continue to reach for yield against a more confident economic backdrop, high yield structured credit may be the natural place to look. From a derivatives perspective, the structured high yield credit market is in its infancy. Yet, from a cash perspective, it is far more developed than even the investment grade market, given the large outstanding and reasonably well-developed secondary market for high yield CBOs and CLOs. In this chapter we focus on understanding this cash and derivatives “gap” and try to predict how the high yield structured credit market might develop going forward. THE FIRST STRUCTURED CREDIT MARKET
The CDO market introduced tranched credit risk to investors, with an almost exclusive focus on high yield bonds and loans. Since 1996, we estimate that over $160 billion of CDOs backed by high yield bonds and leveraged loans have been issued, most of which are still outstanding. The secured nature of many loan transactions, combined with tight covenants and high recovery rates, has resulted in reasonable performance in the CLO sector, compared to a much poorer performance for many high yield CDOs, particularly at the subordinate note level. This early period in the structured credit market was characterized by a lot of new issuance, but remained a fragmented market otherwise with virtually no liquidity and very little transparency. Market participants focused on understanding rating agency methodologies for analyzing default risk. Model-based pricing and correlation risks were terms used by only a few. Investors could go long only, and the “buy-and-hold” nature kept the secondary market from developing. Collateral pools were largely similar in risk profile, centered around single-B rated issuers.
480
WHAT IS DIFFERENT THIS TIME?
To be frank, most of the high yield CBO investors of the late 1990s will never come back to the market, even if it is vastly improved. The pain and career risk they endured during the high yield meltdown was likely too much to warrant giving the market a second chance, at least for now. The new high yield structured credit investor will likely be someone who is similar in profile to the investment grade structured credit investor today: an existing market participant who wants to use long and short tranche positions to help manage a portfolio. What will draw this investor to the market, and what risks will he or she face? LIQUIDITY AND TRANSPARENCY
Despite a much improved secondary high yield CBO/CLO market, it is still difficult for investors to judge valuation and draw comparisons, given the heterogeneous nature of deals. The investment grade market benefits immensely from tranched Dow Jones TRAC-X, where multiple dealers quote two-way markets and liquidity is improving. Can the same thing happen in high yield? Possibly, but it is far from clear right now. While it is natural to think that the market would welcome a benchmark tranched product, we are not certain “who” the market is. Many core high yield investors are not derivatives users. CDO managers could use their “synthetic” buckets for a tranched derivative instrument, but such usage is likely to be small. Hedge funds could be the first interested parties, motivated by the ability to go long and short tranches or employ trading strategies. CDO investors are a possibility as well, but the link between a potential tranched TRAC-X High Yield benchmark (formerly called HYDI) and existing CDOs is not necessarily strong. WHERE DO WE GO FROM HERE? THINK ABOUT BBs
The BB part of the high yield market, from a structured credit perspective, seems intriguing to us. Given the exponential nature of the default rate/credit rating relationship, default risk in BBs is much more tame than the more widely publicized market averages (see Exhibit 1). exhibit 1
Significant Default Experience Difference: BBs vs. Single-Bs
8% 7%
Ba Ba 32 Yr Avg
B B 32 Yr Avg
6% 5% 4% 3% 2% 1% 0% 1970-1975
1977-1982
1984-1989
1991-1996
1998-2003
Source: Morgan Stanley, Moody’s
Please see additional important disclosures at the end of this report.
481
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 69
Over a 32-year period, average annualized five-year BB default rates are 2%, with the distribution ranging from sub 1% levels to just above 4%. For single-Bs, the average was a much higher 5%, with default rates ranging from the 1% level to over 7%. Furthermore, from a fundamental credit valuation perspective, spread per unit of corporate leverage is more attractive in BBs than in any other rating category, as our credit strategist, Greg Peters, has illustrated (see Exhibit 2). exhibit 2
Spread Per Unit of Corporate Leverage Favors BBs
200 180 160 140 120 100 80 60 40 20 0 AA
A+
A-
BBB
BB+
BB-
B
CCCCC
Source: Morgan Stanley
GRAHAM AND DODD VS. BLACK AND SCHOLES
Conceptually, the difference in valuation techniques practiced in the cash and synthetic structured credit markets could not be more stark. Cash players focus on fundamental analysis of the deal, while synthetic market participants use quantitative risk-neutral models. exhibit 3
NAV vs. Derivative Pricing: Convexity Ignored?
Tranche Valuation
7-10% NAV
7-10% Quantitative
95%
75%
55%
35%
15%
-5% Portfolio Spread
Source: Morgan Stanley
482
The fundamental approach of the cash markets involves analysis of many factors, including portfolio NAV, cash flow scenarios, collateral manager quality and structural features. This approach certainly makes sense, given the nature of the cash market, with many deals having highly negotiated structural features and most having managed portfolios. The derivative pricing techniques of the synthetic market rely on the market-implied default probabilities, default correlation and a risk-neutral framework. Valuations generated using these techniques reflect the wider range of possible default scenarios for the portfolio. The result is a price that is more sensitive to small moves in spreads but which may be more stable than that implied by fundamental approaches for large spread moves. This difference leads to what would appear to be inconsistencies between the markets. To illustrate this point, we computed valuations based on both our internal fair value risk-neutral model and purely on portfolio NAV (or liquidation value). Exhibit 3 highlights the difference in valuation of the same instrument, a “senior” tranche of a hypothetical five-year CDO, valued using both a derivative and fundamental approach for a series of average portfolio spreads. While the pure NAV valuation is extreme, it provides a sense for the directional bias of the cash market relative to the derivatives market. All other things being equal, senior tranches tend to be valued lower in the correlation market, while subordinate tranches tend to be valued higher, relative to fundamental approaches. WHERE DO WE START?
Clearly much development needs to happen in the high yield structured credit market before we can think about investment opportunities and relative value, and it is not clear which investors, if any, will “push the envelope.” Yet, the fact that BBs seem intriguing is something that we hope is not ignored by the market place, and we argue that both high yield and investment grade investors should consider the opportunities. Furthermore, a reasonably liquid high yield CBO and CLO secondary market gives us important flow and pricing information, something we never really had in investment grade prior to the past few months.
Please see additional important disclosures at the end of this report.
483
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 70
A View from the Top
March 19, 2004
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Anisha Ambardar Given where we are in the economic cycle and credit spreads, we see both confusion and sticker shock from investors with respect to where senior notes trade. What is the right price for the top part of the capital structure in today’s markets? The cash CDO and synthetic structured credit markets take very different approaches to pricing senior risk, but we would argue that these cultural distinctions are larger than the actual difference in risks. In the structured credit world, the pricing models tell us that super senior risk (typically 10% attachment points for investment grade portfolios) is worth close to nothing in spread terms today. The traditional takers of this risk (reinsurers) are obviously not very excited about getting paid nothing, so they are on the sidelines for the most part. In the cash CDO world, specifically leveraged loan-backed CLOs, top-of-the-capitalstructure senior notes (30% to 40% attachment points) are trading in the mid-40s bp (over Libor) on average in the new issue market, and perhaps 10 to 20 bp wider in the secondary market for healthy deals. Given what we know about super senior risk, are these CLO levels too generous or too tight? We have taken our first steps in applying correlation models (i.e., risk-neutral models of default) to leveraged loan-backed CLOs and find that 30% to 40% subordination in today’s markets translates into extremely little default risk for the tranche. However, CLO structures are far more complex than static synthetic CDOs, so the output of the models should be compared subjectively with the structural and technical aspects of this market, which we do generically in this chapter. In a nutshell, it takes extremely high correlation, combined with extremely low recovery rates for typical top-of-the-capital-structure CLO notes to experience any pricing stress, based on a risk-neutral approach (pricing related to default risk) and today’s market prices for the underlying collateral. From a fundamental perspective, even after adjusting for potential structural risks, we find such senior risks attractive in today’s CLO markets, but caution investors that technical forces may keep the tranches from rallying much further from here. Furthermore, in continuing with our theme of balancing long credit positions with cheap forms of out-of-the-money protection, we recommend that investors consider hedging tail risks in CLO senior notes with long protection positions in 3-100% tranches of investment grade portfolios. In the scenario of extremely high default rates (where investment grade and high yield can be well correlated), this protection will likely be a good hedge against CLO senior pricing stress.
484
CAN CLOs BE QUANTITATIVELY MODELED?
The risk-neutral pricing models that are popular in the structured credit markets are based on the notion that a credit’s default risk is related to the spread at which it trades. In a portfolio context, the distribution of this risk, along with the default correlation of these credits, is used to generate expected values for tranches in a capital structure. In the structured credit world, the structures tend to be simple (static portfolios, no triggers), so the models are reasonable indicators of tranche risk as implied by the underlying markets. CLO structures, on the other hand, are more complicated for many reasons, including the following. •
The underlying portfolios are often not static.
•
There is expected reinvestment of proceeds from maturing and prepaid loans.
•
There are triggers than can pay down tranches and/or delever the structure.
•
The senior notes can get called by equity holders.
CLOs THROUGH A RISK-NEUTRAL LENS
Using the very generic CLO description in Exhibit 1, we take a first look at the “fair value” of a CLO senior note in a risk-neutral framework. The variables that most impact the analysis include the correlation assumption for the portfolio of loans, the recovery rate assumptions and the spread distribution of the underlying loans. exhibit 1 Average Portfolio Spread
Hypothetical Leveraged Loan CLO Senior Note L+287 bp (85% par loans, 15% riskier credits)
Subordination
35%
Average Life
7 Years
Coupon
L + 41 bp
Total Cost of Liabilities
69 bp
Source: Morgan Stanley
Exhibit 2 illustrates the valuation of the tranche in spread terms given changes in recovery rates and correlation levels. Our first insight is that very high correlation, combined with very low recovery rates, is bad for senior notes. The extreme case we modeled would result in a spread of 150 bp (the upper left corner of the surface). However, typical assumptions for both of these variables (30% for correlation and 60% for recovery rates) results in a spread of just 1 bp for the tranche, indicating that the likelihood of loss in the tranche is extremely remote (see Exhibit 3 for the data behind the graph).
Please see additional important disclosures at the end of this report.
485
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 70
exhibit 2
Senior CLO Note Sensitivity to Correlation and Recovery Rates
300 250 Senior Note Spread
200 150 100 50
10% 30%
60 %
80 %
20 %
Recovery Rate
40 %
50% 70%
Correlation
90%
Source: Morgan Stanley
Interestingly, today’s market levels of mid-40s bp for such notes imply correlation/recovery rate relationships that seem unreasonable to us (see shaded regions of table in Exhibit 3). As a benchmark, average historical recovery rates for senior secured high yield loans are in the 60% range, based on Moody’s studies. Implied correlation in the tranched Dow Jones TRAC-X market ranges from about 20% to 35% for tranches today. exhibit 3
Recovery 20% 40% 60% 80%
Correlation 10% 4 1 0 -
The Numbers Behind the Graph – Senior CLO Note Valuation (bp) 20% 14 5 0 -
30% 28 13 1 -
40% 43 22 2 -
50% 60 34 4 -
60% 78 46 7 -
70% 98 61 10 -
80% 121 78 15 -
90% 149 99 22 -
Source: Morgan Stanley
THEORY VS. PRACTICE – STRUCTURAL AND TECHNICAL FORCES
There are several important caveats to the above analysis that are important to consider. First, most CLOs require significant reinvestment of collateral during their life, given short-maturity loans and potential prepayments. All else equal, this reinvestment increases the credit risk of the underlying portfolio. However, if the structure is relatively healthy when the bulk of the reinvestment occurs, then the incremental risk to the senior notes is small, given the remaining term structure and subordination, even if market spreads are wide at the time. A tight spread environment may force the equity holders to call the structure, which could leave senior note holders with reinvestment risk themselves. Second, triggers of any sort are difficult to model, and investors should be concerned about those that can have a negative impact on senior notes, such as any diversion of principal proceeds. Third, the non-static nature of CLOs introduces a variable that is also difficult to model. In today’s credit environment, there is little incremental risk to senior note holders for poor CLO management, but that changes when the credit environment changes.
486
INVESTMENT STRATEGIES
We find good fundamental support for getting long senior CLO risk in today’s markets. However, technical factors certainly limit how much these notes can rally, if any, at this point. We also caution investors to look at the details, in particular subordination levels, as this is one of the biggest factors in any risk-neutral approach to pricing credit risk. For investors who want to hedge the blow-up risk that could lead a CLO senior note to significantly deteriorate, we recommend doing so in places where senior-like risk is priced at tight levels relative to CLOs. One such example exists in the synthetic investment grade market with the now popular 3-100% tranches. Protection on the 3100% tranche of TRAC-X currently trades at roughly 32 bp or 46% of the index spread. While one can question the links between investment grade and high yield performance for marginal changes in spread levels, we argue that the markets are indeed linked in times of abnormally high default rates, which is the true risk to senior notes. One can look at historic instances of extreme default rates for evidence of this link. In past periods of abnormally high default rates, we find high default correlation between investment grade and high yield (see Exhibit 4). Anecdotally, the worst-case 5-year BBB loss rate (assuming 40% recovery in default) is 3.2% since 1970, while for high yield bonds it is 19.5%. Factoring in higher recovery, the implied loss rate for loans would be 13.1%. We point out that the maximum BBB default rate is greater than the 3% attachment point, while the maximum implied loan default rate would eliminate less than half the subordination from the CLO senior note. We believe that 13.1% loss in subordination in a CLO structure (combined with likely wider market spreads) is less painful for a CLO senior note than 3.1% loss in subordination (and wider spreads) is for the 3-100% tranche. This is driven primarily by the time decay benefit that will accrue to the senior note because of the remaining subordination and that will not benefit the 3-100% tranche (under the assumption that losses have significantly reduced subordination). exhibit 4
Aggregate Default Losses Are Well Correlated – High Yield vs. BBBs (5-Year Cumulative)
3.50%
Investment Grade (Left Axis)
25.00%
High Yield (Right Axis)
3.00% 20.00% 2.50% 15.00%
2.00% 1.50%
10.00%
1.00% 5.00% 0.50%
1998
1996
1994
1992
1990
1988
1986
1984
1982
1980
1978
1976
1974
1972
0.00% 1970
0.00%
Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
487
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 71
Origami with CLO Paper?
May 21, 2004
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Anisha Ambardar Angira Apte As the CDO market continues its multi-year evolution, it is clear that what used to be thought of as one investment class has split off into two fairly separate schools of thought, at least from a valuation perspective. In the cash flow CDO world, popular valuation techniques are based on rating agency models and net asset value calculations, while in the synthetic tranched credit world (where the word “CDO” is not used much anymore), correlation-based derivatives pricing models are the norm. In our view, when the underlying collateral is corporate credit, this cultural divide is too stark, as the risks in a cash or synthetic deal are more similar than they are different. With credit fundamentals improving across the board and near-term default risk as low as it has been in years, the new issue CLO market has been very active this year (over $8 billion of issuance, 30% of the CDO market). Over time, CLOs have arguably been the best-performing of the CDO asset types. Today, demand for transactions from good (and even mediocre) managers is pushing pricing to levels not seen since the early days of the market (Libor + 37 bp is the recent tight print for AAA CLO paper). As we have written in previous research (see Chapter 70), when looking at senior CLO notes through a correlation lens, we are quite comfortable with market pricing levels, as there is very little default risk in the top part of the capital structure, resembling super senior tranches in synthetic deals. The next step, though, is to focus on pricing sensitivity to changes spreads, recovery rates and correlation, itself. In CLO mezzanine notes, much like in the synthetic world, these risks are difficult to model and pose some challenges to the conventional wisdom, particularly with respect to portfolio diversity. Varying recovery rate assumptions can lead to opposing sensitivities to correlation, creating an origami-like sensitivity surface for this “paper” and confusing the argument that diversity is a good thing at all levels of the debt capital structure. THE DIFFICULTY OF MODELING MEZZANINE NOTES
In the synthetic tranched credit world, structures are simple, as they rarely have triggers that divert cash flows. Nevertheless, mezzanine notes pose an interesting challenge to the simplest versions of these models, because their pricing can be relatively invariant to changes in correlation values, while their deltas are not (see Chapter 13). Is diversity a good or a bad thing for these tranches? It is not clear, and in fact, our intuition tells us that there must be tranches in the CLO world that suffer from the same fate.
488
THE CORRELATION LENS – FOCUS ON INTUITION
One thing the derivative models can do is give some indication of the relative importance of typical assumptions used by the market in examining valuation for various tranches. In CLOs, recovery rate assumptions are important and can have a significant impact on valuations implied from correlation models. Intuitively, if average recovery rates are high (say 60-80%) then even a small amount of subordination would greatly reduce the risk in junior parts of the capital structure. If the average recovery is markedly lower (20-40%) then the most junior portions of capital structure become much riskier, while the senior portions may only be marginally affected. exhibit 1 Tranche Equity Jr Mezzanine
Generic CLO Capital Structure Assumptions Attachment 0%
Detachment 8%
Size 8% 4%
8%
12%
Mezzanine
12%
15%
3%
Jr Senior
15%
25%
10%
Senior
25%
100%
75%
Source: Morgan Stanley
Historically high recovery rates are a key investment motivation in CLO structures, particularly at the subordinate level. So understanding the risk of unusually low recoveries (independent of actual default rates) is important. In Exhibit 1 we show a generic capital structure and focus on clarifying the correlation (or diversity) sensitivity of the notes. We caution readers that our modeling does not reflect any structural features present in these notes nor do the notes reflect the managed, short-dated nature of the underlying portfolio and the associated reinvestment risk. DIVERSITY SENSITIVITY TO RECOVERY RATES
If we fix recovery rates at a 60% level, which is a popular market assumption, we find that senior tranches’ spreads increase as correlation rises, meaning that they are “short” correlation or “long” diversity. Said another way, a lower correlation portfolio (which translates into higher diversity) lowers the risk in the tranches. This is intuitive and supports the “diversity is good” argument. The two mezzanine notes, however, do not necessarily benefit from higher diversity. The junior mezzanine note’s spread clearly rises as diversity improves (correlation falls) in the portfolio, which implies that an investor in this tranche prefers a low diversity portfolio to one with high diversity. The more senior mezzanine note has a similarly shaped curve, but it is much flatter, and we caution that tranches like this can suffer from the correlation “invariance” problem that we see for 3-7% tranches in Dow Jones CDX today (again, see Chapter 13 for more details).
Please see additional important disclosures at the end of this report.
489
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 71
exhibit 2
Spread
60% Recovery Rates – Diversity Hurts Mezzanine
10%
20%
30%
40%
50% Correlation
60%
Jr Mezz
Mezz
Jr Senior
Senior
70%
80%
90%
Source: Morgan Stanley
If we move to an 80% higher recovery assumption (which is not unheard of in the leveraged loan world), we find that the mezzanine note spread to correlation relationship becomes upward sloping, implying that it behaves just like the senior notes (i.e., diversity benefits the tranche, see Exhibit 3). Wider underlying spreads would work like lower recovery rates, moving the more senior tranches to become “long” correlation in nature. exhibit 3
Spread
80% Recovery Rates – Mezzanine Benefits from Diversity
10%
20%
30%
40%
50%
Jr Mezz
Mezz
Jr Senior
Senior
60%
70%
80%
90%
Source: Morgan Stanley
CLO MEZZANINE NOTE ORIGAMI
As we implied above, conventional wisdom is that diversification within an underlying leveraged loan portfolio is good for the debt investors in a CLO. However, much like what we see in the synthetic investment grade market, with today’s relatively tight spreads, CLO mezzanine notes can be at an inflection point with respect to this rule of thumb.
490
exhibit 4
CLO Mezzanine Note Origami Implied Spread
% 10 % 20
% 50
% 30 % 40
% 70
80%
% 90
60%
% 80
40%
% 60
20% Recovery Rate
tion r el a t i on C or f i ca ersi Div
Source: Morgan Stanley
Although three dimensions are sometimes hard to grasp at first, the “origami” image in Exhibit 4 demonstrates the changing correlation sensitivity for this part of the capital structure. Low recovery rates result in risk rising with diversity, while high recovery rates imply that risk falls as diversity increases. TURNING CONVENTIONAL WISDOM ON ITS HEAD
In today’s spread environment, CLO mezzanine notes reside in the portion of the capital structure that is most sensitive to the lens we used to evaluate them, much like what market participants have learned to accept in the synthetic investment grade markets, using the market’s most basic models. One can think of such mezzanine notes as the untrustworthy ally. When aggregate risk in the portfolio is relatively low, their interests are aligned with senior notes; however, as soon as aggregate risk increases, they switch sides to align with the equity tranche. Either way, the sharper the lens, the better the results.
Please see additional important disclosures at the end of this report.
491
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 72
Seasoning for Convergence: The First Steps
January 11, 2005
Primary Analyst: Vishwanath Tirupattur Primary Analyst: Sivan Mahadevan Ajit Kumar, CFA The emergence of standardized benchmark high grade and high yield credit derivative indices marked a significant turning point in structured credit markets. After an extended period during which several indices were in vogue, credit derivative markets finally came together to standardize the indices. The Dow-Jones CDX IG and DowJones CDX HY indices now trade as market standards for the investment grade and high yield markets for North American credits. Similarly standardized indices now trade for Europe, Japan, Non-Japan Asia and Emerging Market credits. The explosion of liquidity in the trading of benchmark tranches of these indices has added yet another momentous dimension. Expressing refined trading views through correlation trading has become the latest mantra in the structured credit lexicon. The counterpart of the benchmark tranches on the cash side is the already mature but fragmented world of cash CDOs. In principle, corporate credit backed CDOs enable investors to take levered credit exposure on portfolios of corporate bonds and/or loans in much the same fashion as benchmark tranches do. In place of a standardized portfolio of underlying reference entities with the indices, CDO portfolios are largely disparate in their collateral composition. While precise statistics in secondary trading of cash CDO tranches are hard to come by, we note that secondary trading volumes have been surging in the cash CDO tranches. Much has been written about the importance of a well-developed synthetic structured credit market to complement the cash CDO markets. On the surface, the red-hot CLO market, the vintage high yield CBO market and the new-fangled high yield CDX market have a lot in common and the convergence of the synthetic and cash structured credit markets has seemed like a natural progression.
492
WHY THE SLUGGISH CONVERGENCE?
However, the much anticipated convergence has been slow on the uptake. Clearly, there are several material differences between cash CDOs and their synthetic counterparts in the standardized index tranches that explain this apparent sluggishness of the two markets to converge. Cash CDOs have complex waterfall structures that determine payouts to different tranches. Some of these structural features take the form of triggers that determine and change payouts to different tranches according to pre-set rules. Most CDOs have optional redemption features that give the equity tranche holders early call rights. The CDO manager determines the composition of the underlying portfolio subject to trading rules and limitations embedded through various covenants in the CDO governing documents. The portfolios are actively traded for the first few years and become static only at the end of the reinvestment period (typically 3-5 years). For CLOs, the senior secured nature of the leveraged loans renders the delivery of senior unsecured collateral under standard credit default swaps ineffective as hedging instruments. Although the differences in the valuation approaches between the two markets appear to be narrowing as more elaborate models emerge for cash CDO valuation, it is still the case that ratings and static analyses of cash flows under constant default rates with deterministic shocks to key variables play a dominant role in the cash CDO markets. In contrast, synthetic structures are routinely valued under risk-neutral valuation framework with the use of Gaussian copula-type models having emerged as the market standard valuation methodology. SEASONED HIGH YIELD CBOs: A NEW OPPORTUNITY
However, it is now possible to explore the links between some seasoned high yield CBOs and the high yield CDX index to evaluate if some of the barriers enumerated above can be overcome. The deliverable in the event of default in liquid credit default swaps (CDS) is senior unsecured collateral, the same class of collateral as in high grade and high yield bonds. Thus, credit default swaps have emerged as a reasonable instrument for hedging credit risk of senior unsecured bonds, the typical collateral in many seasoned CBOs. Further, several seasoned high yield CBOs, issued about five years ago, are now beyond their reinvestment periods or getting close to the point when trading in the underlying portfolios effectively ceases and the portfolios become static. By evaluating the composition of these now static CBO portfolios and analyzing them with respect to the high yield CDX index, it may be possible to take the first steps towards the convergence of the cash CBOs and synthetic structured credit markets. If there is sufficient overlap in the composition of a particular CBO and the HY CDX index, it may be possible use tranches of the index to hedge existing exposures for investors currently long CBO tranches. With the increased secondary trading in cash CDO tranches, it may further be possible to take long/short positions between the HY CDX index tranches and specific seasoned CBO tranches. We explore this topic further.
Please see additional important disclosures at the end of this report.
493
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 72
exhibit 1 Industry Sector Aerospace/Defense
% of Total 1.00%
Agriculture
1.00%
Airlines
1.00%
Apparel
1.00%
Auto Manufacturers
1.00%
Auto Parts&Equipment
6.00%
Chemicals
7.00%
Commercial Services
3.00%
Computers
1.00%
Electric
5.00%
Electronics
3.00%
Entertainment
1.00%
Environmental Control
1.00%
Food
3.00%
Forest Products&Paper
4.00%
Healthcare-Services
3.00%
Home Builders
3.00%
Insurance
2.00%
Iron/Steel
2.00%
Leisure Time
1.00%
Lodging
4.00%
Machinery-Diversified
2.00%
Media
7.00%
Office/Business Equip
2.00%
Oil&Gas
7.00%
Packaging&Containers
3.00%
Pharmaceuticals
1.00%
Pipelines
3.00%
REITS
4.00%
Retail
5.00%
Semiconductors Telecommunications Source: Morgan Stanley
494
Industry Sector Distribution of HY CDX Index
1.00% 11.00%
A MEDIUM-SIZED UNIVERSE OF SEASONED CDOs
We identified 48 seasoned CDOs whose reinvestment period has either ended during 2004 or will end during 2005. Even though the focus of this chapter is on high yield bonds, we included some CLOs in the analysis since several vintage CLOs have a permissible high yield bond bucket. Using data on collateral holdings of each of these CBOs obtained from Intex, 1 we compare their holdings to the composition of the reference portfolio of the HY CDX index. We identify the % of bond issuers that overlap between each CBO and the index, by number and notional volume. We focus on the tail risk by identifying the bond notional in the CBO rated CCC or below and comparing it to the similarly defined tail risk in the index’s reference portfolio. For each bond issuer in the collateral portfolio, we use the lower of the S&P and equivalent Moody’s rating to determine if the issuer constitutes part of the tail risk. Before discussing the overlap between the collateral portfolios of the CDOs and the reference portfolio of the index, it is useful to examine the composition of the index itself. The reference portfolio of the HY CDX index is an equally weighted static portfolio of 100 bond issuers. Exhibit 1 shows the industry sector composition of the index. Telecommunications is the single largest sector in the index with 11% of the notional. The ratings distribution of the index, using the lower of the S&P and equivalent Moody’s ratings, expressed in S&P ratings terminology is in Exhibit 2. 16% of the index is rated CCC or below. BB and B ratings categories constitute over 80% of the index. exhibit 2
Ratings Distribution of the HY CDX Index
Rating BBBBB+
% of Total 2.0% 9.0%
BB
11.0%
BB-
20.0%
B+
11.0%
B
12.0%
B-
19.0%
CCC+
7.0%
CCC
6.0%
CC
3.0%
Source: Morgan Stanley
Exhibit 3 describes the 48 CDOs analyzed along with their portfolio composition, overlap ratios in the total portfolio and in the tail risk component. Overlap ratios capture the ratio of the bonds overlapping between a CDO’s collateral portfolio and the index’s reference portfolio. 1
As of Dec 23, 2004.
Please see additional important disclosures at the end of this report.
495
496 10/25/05
10/25/00 04/11/01
Titanium CBO I Limited
CANYON CAPITAL CDO 2001-1 LTD.
04/30/05
09/30/04 07/28/04 02/10/05 04/15/04 02/10/05 11/15/05 03/31/04 10/24/04 08/15/04 08/10/05 06/10/04 05/10/05 05/08/05 11/08/05 07/13/05
Issue Date 12/22/98 09/15/99 10/01/00 12/21/99 08/24/00 05/09/00 08/19/99
11/17/99 07/18/00 12/22/99 04/04/00 11/23/99 10/18/01 03/31/99 11/02/99 09/02/99 06/24/99 12/10/99 05/04/00 04/26/00 10/18/00 07/13/00
Name of the CDO Porticoes Funding FMA CBO Funding II Strong CDO III INNER HARBOR CBO 1999-1 LTD. American General CBO 2000-1 Ltd. Gleacher CBO 2000-1 Admiral CBO FREEDOM 1999-1 CDO (PREVIOUSLY CIGNA COLLATERALIZED HOLDINGS 1999-1 CDO) RAINIER CBO I LTD. BATTERY PARK CDO, LIMITED Juniper CBO 2000-1 Ltd. STRONG HIGH YIELD CBO II LTD. Flagstone CBO 2001-1 Phoenix CDO, Limited AIG Global Investment Corp CBO-3 UBS Brinson CBO Limited ML CBO Series 1999-Putnam-1 Cedar CBO CENTURION CDO I, LIMITED Muzinich Cashflow CBO Muzinich Cashflow CBO II Ltd. Wilbraham CBO Ltd.
Comparing CBO Portfolio Composition with DJ HY CDX Index Reference Portfolio
chapter 72
71
60
84%
90%
49 100% 63 100% 47 100% 57 100% 55 100% 77 100% 39 100% 51 100% 52 99% 51 99% 51 98% 58 97% 59 95% 64 94% 55 92% 16%
10%
0% 0% 0% 0% 0% 0% 0% 0% 1% 1% 2% 3% 5% 6% 8% 0%
0%
0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 100%
100%
100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 118
100
75 110 74 87 63 157 43 63 71 109 134 139 68 100 70
25
24
24 37 23 33 25 45 12 25 21 42 50 48 31 30 16
21%
24%
32% 34% 31% 38% 40% 29% 28% 40% 30% 39% 37% 35% 46% 30% 23%
25%
28%
33% 38% 39% 43% 51% 31% 38% 45% 35% 39% 44% 41% 54% 31% 32%
21%
26%
33% 38% 39% 43% 51% 31% 38% 45% 35% 38% 43% 40% 51% 29% 29%
24%
20%
24% 38% 32% 48% 42% 29% 49% 28% 19% 36% 25% 31% 37% 28% 25%
Asset Portfolio Composition # of Unique # of Unique Overlap Ratio (%) Tail Risk Reinvestment WAM % % Bond Issuers Issuers Common # of Bond Total Asset Overlap End Date (Months) Bonds Loans Other Total in CDO to CDO and Index Issuers Notional Notional Ratio (%) 01/22/04 36 100% 0% 0% 100% 54 12 22% 21% 21% 8% 09/25/04 47 100% 0% 0% 100% 47 14 30% 26% 26% 23% 10/15/04 66 100% 0% 0% 100% 85 23 27% 27% 27% 54% 01/15/04 59 100% 0% 0% 100% 124 29 23% 29% 29% 19% 08/24/05 52 100% 0% 0% 100% 101 26 26% 30% 30% 6% 05/09/05 66 100% 0% 0% 100% 98 30 31% 31% 31% 35% 08/12/04 51 100% 0% 0% 100% 36 11 31% 32% 32% 21%
exhibit 3
Structured Credit Insights – Instruments, Valuation and Strategies
Please see additional important disclosures at the end of this report.
497
Comparing CBO Portfolio Composition with DJ HY CDX Index Reference Portfolio
11/05/97 10/22/98 12/15/99 09/16/98 09/14/00 11/14/00 05/04/00 11/02/00 12/16/99 05/13/99 06/24/99 03/15/00 09/15/99 06/22/00 12/08/99 04/15/00 10/19/00
Archimedes Funding Archimedes Funding II Ares III CLO ML CLO Series 1998-Pilgrim America-3 Indosuez Capital Funding VI, Ltd Ares IV CLO Ltd. Stanfield/RMF Centurion CDO II, Ltd. First Dominion Funding III Monument Capital Stanfield CLO, Ltd. Longhorn CDO (Cayman) Ltd SIMSBURY CLO, LIMITED AIMCO CDO, Series 2000-A Octagon Investment Partners III, Ltd AMMC CDO I ADDISON CDO, LIMITED
11/08/04 10/08/05 01/12/05 09/23/04 09/14/05 12/22/05 04/15/05 11/12/05 12/22/04 05/13/04 07/15/04 05/10/05 09/24/04 08/22/05 12/15/04 01/15/05 11/08/05
11/15/04 12/19/05 12/12/04 05/11/04 01/13/05
Source: Intex Solutions Inc and Morgan Stanley calculations
11/09/00 12/19/00 12/02/99 05/11/99 12/13/01
Name of the CDO SAAR HOLDINGS CDO, LIMITED Magma CDO (previously Whitney Cash Flow Fund II) Blue Eagle CDO I S.A. Catalina CDO Bedford CDO, Limited Ecureuil Credit Plus CDO 31 33 64 46 72 63 65 64 60 51 58 61 62 64 61 61 66
73 64 67 52 78 40% 34% 30% 27% 27% 25% 23% 22% 22% 22% 21% 20% 20% 17% 15% 15% 11%
74% 72% 70% 60% 60% 60% 66% 70% 73% 73% 75% 77% 78% 78% 78% 79% 80% 80% 83% 85% 85% 89%
26% 27% 30% 40% 33% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0%
0% 1% 0% 0% 7% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100%
100% 100% 100% 100% 100% 13 13 62 20 34 72 66 121 72 39 43 25 43 86 56 10 25
67 39 62 60 63 0 2 17 8 13 19 19 32 15 19 12 10 9 17 26 4 11
17 4 21 23 7
0% 15% 27% 40% 38% 26% 29% 26% 21% 49% 28% 40% 21% 20% 46% 40% 44%
25% 10% 34% 38% 11%
0% 10% 29% 35% 44% 32% 28% 27% 21% 44% 35% 61% 20% 20% 54% 40% 45%
32% 13% 35% 38% 9%
0% 3% 8% 9% 12% 8% 7% 6% 5% 10% 7% 12% 4% 3% 8% 6% 5%
23% 10% 25% 23% 5%
0% 0% 38% 25% 32% 32% 20% 16% 36% 56% 25% 63% 13% 26% 76% 84% 56%
27% 40% 55% 46% 4%
Asset Portfolio Composition # of Unique # of Unique Overlap Ratio (%) Tail Risk Reinvestment WAM % % Bond Issuers Issuers Common # of Bond Total Asset Overlap Issue Date End Date (Months) Bonds Loans Other Total in CDO to CDO and Index Issuers Notional Notional Ratio (%) 03/16/99 03/24/04 49 78% 22% 0% 100% 63 12 19% 24% 19% 6%
exhibit 3 (cont.)
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 72
AN ILLUSTRATIVE EXAMPLE
To illustrate, consider Strong High Yield CBO II Ltd highlighted in Exhibit 3. Issued originally in November 1999, the CDO’s reinvestment period ends in February 2005. The weighted average maturity of the assets in the portfolio is 55 months (4.6 years). The collateral portfolio of the CDO consists of 100% high yield bonds with 63 unique bond issuers. 25 of the bond issuers in the CDO’s collateral portfolio are also in the index’s reference portfolio. Unlike the index, the bonds in the CDO’s collateral pool are not equally weighted. In notional terms, about 51% of the bonds in the portfolio overlap between the CDO and the index. In terms of tail risk, 42% of the bond notional in the tail risk component of the portfolio overlaps with the index’s reference portfolio. In Exhibits 4 and 5, we compare the non-overlapping components of the index and the CBO in terms of industry sectors and ratings. The mismatch between the index and the CBO in the non-overlapping portion is significant. The non-overlapping portion of the CBO is weighted more heavily in the CCC ratings category than the index is. While the industry sectors also show a mismatch, the extent of the mismatch can be evaluated by sector specific views.
498
exhibit 4
Sector Distribution of Non-Overlapping Notional
Industry Sector Commercial Services
Strong CBO II 11.5%
DJ HY CDX Index 4.0%
Media
9.2%
5.3%
Retail
8.5%
4.0%
Lodging
7.9%
5.3%
Chemicals
6.3%
5.3%
Office/Business Equip
5.9%
1.3%
Entertainment
5.9%
0.0%
Office Furnishings
5.1%
0.0%
Telecommunications
4.6%
9.3%
Advertising
4.2%
0.0%
Iron/Steel
4.1%
1.3%
Cosmetics/Personal Care
3.9%
0.0%
Distribution/Wholesale
3.4%
0.0%
Leisure Time
3.4%
0.0%
Trucking&Leasing
2.5%
0.0%
Healthcare-Services
1.9%
4.0%
Packaging&Containers
1.7%
4.0%
Electric
1.7%
5.3%
Oil&Gas Services
1.7%
0.0%
Pharmaceuticals
1.7%
1.3%
Diversified Finan Serv
1.7%
0.0%
Electronics
1.7%
2.7%
Forest Products&Paper
1.5%
4.0%
Aerospace/Defense
0.2%
1.3%
Environmental Control
0.0%
0.0%
Semiconductors
0.0%
0.0%
Auto Parts&Equipment
0.0%
5.3%
Apparel
0.0%
0.0%
Oil&Gas
0.0%
8.0%
REITS
0.0%
5.3%
Home Builders
0.0%
4.0%
Food
0.0%
4.0%
Pipelines
0.0%
4.0%
Machinery-Diversified
0.0%
2.7%
Insurance
0.0%
2.7%
Airlines
0.0%
1.3%
Auto Manufacturers
0.0%
1.3%
Agriculture
0.0%
1.3%
Computers Total
0.0%
1.3%
100.0%
100.0%
Source: Intex Solutions Inc and Morgan Stanley
Please see additional important disclosures at the end of this report.
499
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 72
exhibit 5 Rating BBB
Ratings Distribution of Non-Over Lapping Notional Strong CBO II 1.69%
DJ HY CDX Index 0.0%
BBB-
0.00%
2.7%
BB+
0.00%
12.0%
BB
0.00%
12.0%
BB-
10.66%
21.3%
B+
4.23%
8.0%
18.10%
10.7%
B-
17.43%
21.3%
CCC+
18.13%
6.7%
CCC
8.46%
5.3%
CCC-
11.16%
0.0%
5.07%
0.0%
B
CC C Total
5.07%
0.0%
100.00%
100.00%
Source: Intex Solutions Inc and Morgan Stanley
What does this example illustrate? Roughly speaking, about half of the credit risk in the CBO can be hedged by taking an offsetting position in the index. Assuming that issuers in the tail risk, as defined here, constitute issuers most at risk for default, 42% of the tail risk can also be hedged with the index. Even though there is not a 100% overlap between the index and the CBO, it might still be possible to explore trading opportunities between the two. We will argue that to take offsetting positions in the mezzanine tranches of the CBO and the HY CDX index an overlap of 100% is probably not necessary. Remember that the risk exposure of a mezzanine holder is not to the default of the entire portfolio but losses in a defined range. The further up in the capital structure, the less is the sensitivity to the overlap in the tail risk portion. In contrast, lower down the capital structure, tail risk overlap is more material. Obviously, tail risk overlap matters more to equity tranches than to mezzanine tranches. While there remains a basis risk, it may be possible to hedge the non-overlapping component with single name credit default swaps. It seems to us that at the minimum there is a potential to further explore trading/hedging strategies between a mezzanine tranche of the Strong High Yield CBO II and an appropriate tranche of the HY CDX index. More generally, our analysis shows that the overlap ratios vary widely – between 0% and 51%, with a majority of cases in the 30-45% range. Tail risk overlap ratios also cover a similarly broad range. Where there is a significant overlap, a potential exists to explore the relationships further which may offer trading and/or hedging opportunities. Several market participants are already using the single-name CDS market to hedge specific exposures in CBO tranches on a delta–adjusted basis. We argue that application of HY CDX index and the tranches of the HY CDX index is the next natural progression in the path towards convergence of cash and synthetic markets.
500
NEXT STEPS
Clearly, this is only a first step in the analysis. A more complete analysis should include further examination of tranche analytics to determine the index tranche most appropriate for exploiting the relationships between the two markets. Our definition of tail risk can be further refined by adding the level of spreads to the analysis. Besides the commonality of credit exposure, there are other issues that need to be evaluated. Included among issues not explored here but deserve to be analyzed are: the option that the CBO equity holders have, to force an early redemption of the debt tranches and the unequal weighting of individual issuers in the CBO’s underlying portfolio. The valuation divide between synthetic and cash CDO markets implies that there is a potential for mark-to-market risk when taking offsetting positions in these markets, even when the overlap of exposures is perfect. We do not intend to make light of these complexities. Notwithstanding such challenges, this analysis marks our first steps in the exploration of the convergence between the cash CDOs and the synthetic structured credit markets. We are sure to revisit this issue in the future. Watch this space.
Please see additional important disclosures at the end of this report.
501
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 73
Cash & Synthetic CDOs – So Where’s the Convergence?
December 2, 2005
Primary Analyst: Sivan Mahadevan Primary Analyst: Vishwanath Tirupattur Primary Analyst: Peter Polanskyj There is no simple answer to the often-asked question about why cash and synthetic structured credit markets are so different. We have argued in the past that the nature of the ultimate credit risk on both sides of the cash and synthetic divide are largely the same, but structural features, trading practices, common collateral types and valuation culture continue to be more different than similar (for our early thoughts, see Chapters 69 through 72). For those who didn’t live through all of the history, it may be useful to highlight that the cash and synthetic CDO markets were one and the same over the first several years of this now 10-year-old market. The rapid growth of credit derivatives, their indices and tranches in the corporate credit space, the flexibility and ease of creation of static synthetic CDOs, the liquidity and transparency of a benchmark and a derivatives pricing culture created a different kind of market on the synthetic side. Meanwhile, the cash side has seen less transformative change, but has experienced an explosion of growth, with underlying collateral classes largely bereft of (until recently) synthetics, and has found a natural home in the hands of securitized products buyers. Nevertheless, there are convergence themes, as relatively powerful outside forces like Basel II, along with increasing idiosyncratic credit risk are forcing shifts in valuation. In this chapter, we focus on where the markets have converged and how they are different, as well as examine potential future convergence paths. WHERE HAVE THEY CONVERGED?
The boost in liquidity, transparency and associated market flows that the synthetic side of the structured credit market has enjoyed has been a tremendous source of curiosity for all credit investors, whether or not they participated. It was quite easy for many to predict that cash CDO valuations would eventually follow the correlation model path, but ironically, in 2005, quite the opposite has been true. The sharp rally in synthetic mezzanine and senior tranches, combined with somewhat wider spreads and a general sell-off in equity tranches has created an extremely large correlation skew even in the face of falling implied correlation for equity tranches (see Exhibit 1).
502
In many ways, pricing in the senior part of the market is more closely related to drivers of cash CDO prices, where ratings and perceived ratings stability are important factors. We explored this concept in greater detail recently (see Chapters 52 and 55). In particular, senior tranches get beneficial treatment under Basel II, impacting both cash and synthetic tranches; even recent FASB proposals potentially put funded synthetics on more equal footing with cash CDOs from an accounting perspective for US insurance companies and banks. exhibit 1
60%
Repricing Synthetic Senior Tranches
50%
A BBB
40%
30%
20%
10%
2000
1998
1996
1994
1992
1990
1988
1986
1984
1982
1980
1978
1976
1974
1972
1970
0%
Source: Morgan Stanley
Although it took a direction that many did not predict, the above phenomenon is a significant convergence theme. We would add that ratings agencies are increasingly approaching the rating of both cash and synthetic CDOs with a similar modeling framework. S&P, for example, uses its CDO Evaluator model to determine the probability distribution of potential default rates for both cash and synthetic CDOs. And although Moody’s still uses somewhat different approaches for rating synthetic corporate CDOs and cash CLOs, its modeling approach is increasingly converging between corporate and ABS CDOs. TODAY’S CONVERGENCE PLAYS – SMALLER OPPORTUNITY
Where can investors play convergence themes today? One of the biggest differences between cash and synthetic CDOs is in collateral type, which makes it difficult to play convergence themes against one another, except perhaps at senior parts of the capital structure. Most of the growth in cash CDO issuance over the past three years has been in ABS CDOs and CLOs, the two CDO sub-sectors with little activity on the synthetic front (while ABS synthetics have started to take off after the standardization of documentation in June 2005, the depth and liquidity is still evolving). Further, no standardized index (or index tranches) are actively traded in loans or ABS. This is important because the presence of such traded instruments provides key benchmark price and correlation observations. We argue that the absence of such instruments has limited convergence between cash CDOs and the structured credit market.
Please see additional important disclosures at the end of this report.
503
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 73
The one area where convergence is the most natural is in high yield synthetic CDOs (most of the activity lies in the 5-year HY CDX tranches) relative to seasoned HY CBOs, many of which are in the 5-year maturity neighborhood today and have exceeded their reinvestment periods. We explored this opportunity in detail almost a year ago, identifying several situations where seasoned HY CBOs had portfolios with a fair amount of overlap with the HY CDX index (see Chapter 72). There has indeed been more hedging activity between HY CBO tranches and the HY CDX index over the past year, but the amount of absolute secondary trading activity in HY CBOs has decreased due to redemptions and deleveraging. CLO CONVERGENCE IS TEMPTING
It is tempting to apply risk-neutral correlation model approaches to valuing CLOs, and we have done so in the past, knowing fully well the shortcomings of not modeling a waterfall (see Chapters 70-71 as well as Chapter 50). However, we still find the relative shifts in model-predicted valuations to be quite insightful. In Exhibit 2, we show how typical leveraged loan portfolios (within CLOs) have changed at different points this year and then compare observed shifts in CLO tranche spreads to modelpredicted shifts. The correlations we used were calibrated to make the first period of pricing equal to what was observed in the market. In general, models would have pushed BBB, A, and AA tranches wider in May with the underlying market, although in index tranches, actual price moves similarly disagreed with models. exhibit 2
CLO Rated Tranche Price Movements – What Do the Models Say? (bp) Jan-05
May-05
Nov-05
BB
213
204
167
B
259
286
265
Second Lien
571
530
525
Wt. Avg CLO Asset Spread
285
301
279 185
Loan Spreads
Actual CLO Spreads BBB
200
185
A
95
75
75
AA
56
42
45
Model Predicted CLO Spread Difference BBB
+55
0
A
+47
+7
AA
+31
+4
Source: Morgan Stanley
504
WHERE HAVE THEY BECOME MORE DIFFERENT?
The synthetic structured credit market gets broader attention in part because of the liquidity and transparency of the index tranche market and the advent of pricing models. This leads to more rational relative pricing among opportunities in all market environments, with 2005 being a good example. This model is not necessarily transferable to the cash CDO world, where structures are more complicated (waterfalls, amortization, reinvestments and optional redemption features), and ratings and subjective measures like managers are important pricing inputs. A natural buyer base for cash CDOs has developed, most notably in securitized products portfolios. Despite the benefits of liquidity and transparency, synthetic structured credit products do not enjoy the same natural buyer base, even though a broad base of investors participates. The complexity of cash CDOs relative to synthetics makes them difficult to price through stochastic methods, such as Monte Carlo simulations that are the bread and butter of correlation models. Random portfolio losses in cash CDOs can impact tranches in very different ways depending on when and how they actually occur, due mostly to the highly engineered cashflow waterfalls. Perhaps the easiest tranches to compare in both worlds are equity tranches, but even here we see some large cultural differences. Equity tranches in synthetics trade on a price and correlation basis where views on spread dispersion and implied correlation are big drivers of opportunities. But synthetic equity tranches are for the advanced user, mainly because most are static in nature and generally held both on a delta-adjusted basis and where ultimate gains are related to the success of a jump-to-default hedge strategy. Equity tranches in cash CDOs are viewed more on an IRR basis (much like investments in other worlds like distressed debt and private equity). Investors tend to assess risk through non-stochastic default paths (for example, a 2% constant annual default rate). For those who try to “price” both opportunities in a risk-neutral sense, synthetic equity tranches appear cheaper in today’s environment, thanks in part to the big repricing that has occurred this year. But the waterfall in cash CDOs, even for equity, can be valuable in some cases, as is the call option.
Please see additional important disclosures at the end of this report.
505
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 73
CALL ME, IF YOU CAN
In fact, the significance of the call option may be seen in the dramatic pick-up in the optional redemption activity we have noted in cash CDOs (see Exhibit 3). Tight spreads and a relatively low interest rate environment have created ideal conditions for the call options embedded in the equity tranches of cash CDOs to be significantly in the money. exhibit 3
CDO Optional Redemptions
# of Deals 16 14 12 10 8 6 4 2 0 2003 Q2
2003 Q3
2003 Q4
2004 Q1
2004 Q2
2004 Q3
2004 Q4
2005 Q1
2005 Q2
2005 Q3
2005 Q4
Note: 2005 Q4 numbers are as of end of October 2005. Source: Morgan Stanley, Moody’s, S&P and Fitch IBCA.
While the exercise of these call options has certain built-in inefficiencies, the huge pick-up in call activity suggests that it can be valuable. Yet, valuation of such options in a consistent risk-neutral fashion is challenging. This is not an issue that the structured credit market has to contend with. WHO ARE THE INVESTORS?
Another source of difference is the investor base. Cash CDOs, including CLOs and SF CDOs, are generally used by investors to outsource credit expertise, thereby gaining exposure to markets that may not be available to them directly. The buyer base includes banks (largely at the senior level but also at the mezzanine level), insurers and, ironically, CDO managers of other asset classes (i.e., SF CDO managers buying CLO tranches and vice versa). The exception to this trend is the fairly significant amount of CDO risk held by managers. The synthetic CDO market, including all of the index tranches, has perhaps a broader client base because both credit pickers and credit outsourcers are involved, as are those who have a preference to take exposure in derivative form, from re-insurers to hedge funds. There is a bit of a renaissance in managed synthetic CDOs today, given structural features that make it easier for dealers to hedge the residual risks and changing portfolios. Such structures have much more in common with traditionally managed cash CDOs, from a buyer base perspective.
506
THE FUTURE PATH OF CONVERGENCE
There will be some interesting convergence paths between cash and synthetic CDOs in the not-so-distant future, in our view. These will be driven both by the development of synthetics in the leveraged loan and structured finance space and the increasingly negative credit environment in both corporate and potentially consumer credit. We have addressed in previous research both the development of standardized credit derivatives in the structured finance space and the emerging prominence of pure or hybrid synthetic structured finance CDOs. While synthetic tranches from this world do not really trade on any model based pricing approach, this may become more popular going forward, particularly as the stress that has hit the weaker parts of the HEL sector unfolds. On the corporate credit side, the development of leveraged loan CDS could create an interesting synthetic alternative to the much heralded cash CLO market (more so in Europe than in the US for now – see Viktor Hjort’s report, “Leveraged Loan CDS: The Final Piece of Jigsaw,” November 4, 2005). If it occurs, this would be the first time that healthy “new issue” activity in both cash and synthetic corporate CDOs exists for the same type of collateral. Pricing comparisons will then be easier to make. Independent of leveraged loan CDS development, as a more negative credit environment undoubtedly impacts the red hot leveraged loan market, CLOs will begin to feel some pain; this would again bring up the valuation question, prompting comparisons with synthetic structured credit (see Chapter 56). In any event, while the convergence of cash and synthetic structured credit opportunities has been much slower than many had predicted, there are indeed parts of the market that are ripe for convergence, but their key differences will remain significant.
Please see additional important disclosures at the end of this report.
507
Section G
FTD Basket Investment Themes
Please see additional important disclosures at the end of this report.
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 74
Levering the Lightly Levered
September 12, 2003
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Anisha Ambardar Leverage may be an overused term in financial markets, but it is a word that aptly describes how investors or corporate management can “re-size” an investment or enterprise, thereby changing its risk and return profile. In credit markets we are all used to thinking about “corporate” leverage (debt/EBITDA), but tranched basket products allow investors to add “structural” leverage as well. The “total” leverage resulting from this exercise is an interesting concept to consider, and we argue that investors hunting for yield in this market should combine these leverage concepts and compare resulting opportunities. In this chapter we focus on identifying investment ideas based on both of these notions of leverage, using first-to-default baskets as the structural vehicle. GETTING PAID FOR CORPORATE LEVERAGE?
Corporate leverage metrics, which we define as outstanding debt divided by an earnings measure like EBITDA, are an important driver of credit risk and as such are key inputs in the analyst rating process (used both by credit analysts and rating agencies). Greg Peters, our chief credit strategist, has combined corporate leverage with spreads to create relative value tools in both investment grade and high yield markets. Broadly speaking, corporate leverage rises with decreasing credit quality, but the shape of this curve is not necessarily consistent with market spreads (see Exhibit 1, where results are based on second quarter balance sheet data from 600 non-financial issuers). Based on Greg’s research, for direct credit investors, the “sweet spot” in spreads versus leverage currently is in the high-quality space (AAA/AAs) and in BBs. As a caveat, we remind readers that corporate leverage is not the only driver of credit risk, although it is an important one. “TOTAL” LEVERAGE OPPORTUNITIES
Structured credit investors can use this valuable corporate leverage information in two important ways. First, on a “total” leverage basis, levering up AAAs and AAs (where ultimate ratings could be AA to A) may be better than buying single-As and BBBs outright since resulting structures may offer more spread per unit of “total” leverage than the outright investments. The same argument could be made for levering up BBs versus buying single-Bs outright.
510
exhibit 1
The Sweet Spots: Spread Per Unit of Corporate Leverage
Spread Per Unit of Debt/EBITDA 200
Debt/EBITDA 9.00 8.00 7.00
180
Debt/EBITDA Spread Per Unit of Leverage
160 140
6.00
120
5.00
100 4.00
80
3.00
60
B
B-
B+
BB BB -
BB B BB BBB +
A -
BB B+
0 A
20
0.00 A A A A A +
40
1.00
A A A
2.00
Source: Morgan Stanley, Factset
A review of Moody’s historical cumulative 5-year default data and current spread levels provides some further insight on this front. We introduce another relative value measure – spread per unit of default risk 1 – in order to reflect the impact of both corporate and structural leverage. When we compare the average spread per unit of default risk of a levered AAA/AA basket against that of an average BBB credit, we find (under the conservative assumption of zero default correlation) that the levered product offers approximately 2.4x the spread per unit of default risk. Additionally, a similar analysis for a levered BB basket results in 1.5x the spread per unit of default risk when compared to an average B credit. Our second argument for structured credit investors is that, while the data in Exhibit 1 is based on average numbers, the outliers in these distributions are interesting credits as well, from a long or short perspective. In Exhibit 2, we show the distribution of corporate leverage numbers for 265 single-A and BBB issuers. The leverage numbers range from near zero to above 10x, with the average being 2.7x. Selling first-to-default protection on some of the lightly levered names in this universe may be a good relative value trade for investors with positive or neutral views on the credits otherwise. TAKING THE CONCEPT FORWARD: SMALL BASKET IDEAS
First-to-default baskets are good vehicles for credit pickers to take on incremental, name-specific credit risk. Based on the flows we have seen this year, first-to-default basket usage is quite different from usage in the product’s bigger cousin: single tranches of large portfolios including TRAC-X. Ignoring the hedge fund universe for a moment, we have found flows in the small-basket market tilted toward large, creditsavvy real-money investors who welcome the opportunity to implement credit views and often look for such opportunities away from the core part of the investment grade market, where they generally already have enough exposure. Flows in the large-basket products are probably more evenly distributed among all credit investors, given that exposure to large issuers and diversification are welcomed in many portfolios. 1
Defined as credit spread/annualized 5-year default probability.
Please see additional important disclosures at the end of this report.
511
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 74
exhibit 2
Relative Value Dispersion in A/BBB Corporate Leverage
Frequency 45 40 35 30 25 20 15 10 5 0 1
2
3
4
5
6
7
8 9 10 11 12 13 14 15 16 17 18 19 20 Debt/EBITDA Buckets
Source: Morgan Stanley, Factset
With this background in mind, we have designed three first-to-default baskets that we believe are well-suited for investors confident enough to take on some additional credit risk by levering up attractively levered companies today. While we think the baskets themselves are interesting, the methodology behind creating them is probably even more important. LEVERING HIGH-QUALITY NAMES
First-to-default baskets built from AAA and AA credits have been the most popular flow in this small market this year, with real-money investors being the natural sellers of protection. The idea of levering up lightly levered names is appealing, both from a credit risk perspective and from a diversification perspective, given that most large credit investors have adequate exposures to the core A and BBB parts of the corporate markets. exhibit 3 High Quality Underlying Credits
Mid
BBB Outliers
Mid BBs
Mid
Home Depot
18
Sprint
110 D.R. Horton
160
UPS
13
EDS
220 Nextel
255
Merck
13
Delphi
160 Lear
155
SBC Comm.
40
Altria Group
332
IBM
33
Avg Debt/EBITDA
1.7x
0.9x
2.9x
FTD Prem
Bid 85
Offer 103
Bid 613
Offer 643
Bid 405
Offer 495
% of Total
73%
88%
75%
78%
71%
87%
Correlation
60%
35%
44%
37%
61%
30%
Source: Morgan Stanley
512
Levering the Lightly Levered Credits
Our benchmark high-quality basket (comprised of FNMA, AIG, GECC, PFE and WMT) has an average default swap premium of 27 bp today, 10 bp tighter than its level in early February, the wide of this year’s spreads. The bid side of the first-todefault basket, however, has tightened from a 130 bp level to 96 bp today. Implied correlation is high, given our subjective view that these credits are not very correlated, but today’s implied correlation level of 55% is lower than the peak level of 60%. The high-quality basket we show in Exhibit 3 is different in nature from our benchmark basket. It includes five credits that are all non-financial in nature and have an average debt/EBITDA ratio of 0.9x, well below the average level of 1.5x for AAs. The seller of first-to-default protection on this basket receives 73% of the total premium, with an implied correlation of 60%. HIGH YIELD
On an absolute basis, the BB sector appears attractive when looking at the spread per unit of corporate leverage metric, but an investor must be comfortable with a levered investment in BB credits. We have selected a basket with three names and are thus applying less structural leverage than we do above. Investors could add names to generate more structural leverage and achieve a more single-B-like “total” credit quality. The names in this basket are somewhat tighter than typical BBs, but the basket’s average corporate leverage is significantly lower than the average BB measures (2.9x vs. 4.4x). The resulting spread on the basket is 405 bp, with implied correlation of 61%. OUTLIERS IN A CORE MARKET
While average spread per unit of corporate leverage numbers appear less attractive in the BBB sector, dispersion in leverage among the issuers is a good source of relative value, as we highlighted in Exhibit 2. Our “BBB Outliers” basket in Exhibit 3 was constructed as a result of this dispersion and contains four BBB credits with average corporate leverage of 1.7x versus 3.1x for BBBs as a whole. The spread on the first-todefault basket is 613 bp, where ultimate default risk may be BB-like. With this basket in particular, investors need to be comfortable with the non-leverage-related factors affecting credit quality in the underlying names.
Please see additional important disclosures at the end of this report.
513
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 75
Managing Merger Mania
October 31, 2003
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Anisha Ambardar In a credit market that continues to be amazingly tight and compressed, a pickup in merger activity is introducing both volatility and opportunity. For credit investors wondering how they are going to make money in 2004, merger activity may be one area that merits their attention. At the single-name level, understanding mergers is all about credit analysis. Balance sheets, debt ratios and credit ratings are the key areas requiring focus. Investors with keen insight into the resulting impact are the ones who will be rewarded first, if fundamentals eventually play out. Mergers have an interesting impact on first-to-default baskets, and we feel this effect is not well understood in the marketplace. Depending on the nature of “replacement” language in contracts, sellers of protection on a first-to-default basket are effectively long a merger option on the names in the basket. This merger option does not require the investor to pick the specific credits that could merge; the candidate credits simply need to be included in the basket. Investors can thus focus on a sector or group of credits where merger activity is a possibility and get rewarded when an actual merger occurs. By contrast, in the single name world, one has to pick the right credits and also correctly discern the direction of spreads to benefit. We focus this chapter on better understanding the merger options inherent in small baskets and are fortunate to be able to track valuations in a real-life example involving our Benchmark Bank Basket. exhibit 1
Can We Expect More Merger Activity?
Other Food & Consumer Prods. Retail Utilities Natural Resources Healthcare TMT Financial Services
$Billions 3,000
2,500
2,000
1,500
1,000
500
0 1998
1999
2000
2001
2002
2003 YTD
Source: Morgan Stanley, Thomson Financial
MERGERS AND CORRELATION
In the context of baskets and tranches, a merger between two reference entities is equivalent to an upward jump in average correlation (because the correlation of the two names goes to 1). Such a jump can have a dramatic impact on the valuation of small baskets. Those who are long correlation (e.g., sellers of first-to-default protection) will
514
benefit when a merger occurs, and often the benefit can more than offset a potential widening of spreads, if the merger is not bondholder friendly. Let’s focus on a specific example: On October 27, Bank of America and FleetBoston Financial announced a merger. In the default swap market, Fleet traded about 8 bp wider than Bank of America prior to the announcement; afterwards, both traded at or near the original Bank of America level (of about 20 bp mid-market). exhibit 2
BAC and FBF Merger – FTD Basket Jumps 10 bp
FTD Bank Basket
Before Merger Announcement 67/80
After Merger Announcement 57/67
Total/Avg Spread
116/23
108/22
BAC
20
20
C
20
20
FBF
28
20
ONE
22
22
WFC
26
26
Source: Morgan Stanley
Our Benchmark Bank Basket has a total of five banks, including both Bank of America and FleetBoston. The basket trades without replacement language (which we will discuss below). As such, from the perspective of a first-to-default buyer or seller, it will effectively be a four-name basket going forward, if the announced merger goes through. Going from five to four credits can be considered a reduction in structural leverage (see Chapter 74), which benefits the seller of first-to-default protection (because the investment becomes less risky). Based on market pricing, the basket’s premium rallied 10 bp (from 67 to 57 bp on the bid side), purely as a result of the announced merger (spreads on the other banks did not move). THE MECHANICS – BASKETS WITHOUT REPLACEMENT LANGUAGE
The mechanics behind merger treatment in baskets are important to understand, as there are two types of language that trade in the market (see Exhibit 3). In the simplest form (i.e., without replacement language), a five-name first-to-default basket where two reference entities merge effectively becomes a four-name basket once the merger closes. However, the “contractual” notional amount of the merged entity in the basket doubles to reflect that it was once two reference entities. Yet, this change in exposure size of the merged entity does not matter to the seller or buyer of first-to-default protection. A credit event in any of the four names (including the merged entity) would still trigger the contract. A credit event experienced by the merged entity, however, has a larger notional impact. Therefore, it would simultaneously trigger both a first-to-default and a second-to-default contract.
Please see additional important disclosures at the end of this report.
515
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 75
exhibit 3
Millions 25
Replacement Language Matters – Impact of a Hypothetical Merger Between Credits 1 & 2
20 15 10 1
5
2
0
3
4
Credits
No Replacement
Replacement
Millions 25
Millions 25
20
20
10 5 0
6
15
15 1 & 2
5
2
10 3
4 Credits
5
5 0
1
3
4
5
Credits
Note: Assumes $10 million notional of first-to-default exposure. Source: Morgan Stanley
REPLACEMENT LANGUAGE – OPERATIONALLY PUZZLING
The impact of a lack of replacement language seems intuitive to us, given either the correlation or leverage reduction analogies we described above. The seller of first-todefault protection sees a reduction in risk when there is a merger, so the seller of second-to-default protection must see an increase in risk, for which he or she must be compensated. Market clearing levels should take this into consideration, if the market is efficient about this. In fact, we would argue that the merger option is one factor that keeps implied correlation in small baskets higher than in large baskets. First-to-default baskets trade with “replacement” language in the marketplace as well (see Exhibit 3). Here, when a merger occurs, instead of dropping an entity in the basket, the buyer and seller of protection must mutually agree on a replacement credit, and often this replacement credit is one that trades at a similar spread level. Clearly, there is significant correlation risk in this exercise, and the buyer and seller of protection will have opposite correlation views. Furthermore, the operational aspects of going through this process for each basket that experiences a merger can be tedious. We have much more experience with baskets that trade without replacement language, which again seems more intuitive to us. We thus recommend that both buyers and sellers of first-todefault protection avoid replacement language. THE MERGER IMPACT – WHAT IS IT WORTH?
The 10 bp impact in pricing that we saw for our Benchmark Bank Basket may not seem like a lot of spread movement for being long a merger option and getting it right. However, we do consider it a reasonable move on a relative basis, given the nature of the basket. The average premium of the five banks was 22 bp, and the basket was priced at a relatively high correlation already (76%, on the bid side) before the merger announcement.
516
What Is the Impact of a Merger Worth?
exhibit 4
Average Spread of Credits in Basket (bp) 50 100 41 78
Correlation 20%
20 18
150 111
40%
15
32
57
80
60%
11
22
39
53
80%
6
13
22
29
Source: Morgan Stanley
In order to better understand the value of the merger options, we examine the pricing impact across credits with a variety of spreads and correlations. Generally, the merger impact becomes more valuable for riskier (e.g., higher spread) credits, as well as for baskets where implied correlation (before the merger announcement) is low. For example, the expected spread move in a basket with an average premium of 50 bp, at a correlation of 40% (which is typical for diversified industrial first-to-default baskets), would be 32 bp (see Exhibit 4). At 150 bp of average premium, it would be worth 80 bp. This assumes no movement in spread of the merged entity or other credits in the basket. To the extent that a merger drives the merged entity’s spread wider, the impact of the merger would be more muted than indicated in the table. As an example, for a basket with average spread of 50 bp, trading at 60% correlation, the gain on the merger of two entities (roughly 22 bp of first-to-default premium) would be fully offset by a 33 bp (66%) widening in spread for the merged entity. WHAT ARE THE TRADES TO DO?
It depends on one’s view of where the merger activity is going to be. Historically, financial services and TMT have been popular sectors (see Exhibit 1), given both consolidation and growth opportunities. Our main point is that the first-to-default basket market is not yet sophisticated enough to reflect market merger views, so investors may find some interesting merger option opportunities.
Please see additional important disclosures at the end of this report.
517
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 76
Levering the Fundamentally Fit
April 2, 2004
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Anisha Ambardar Credit fundamentals continue to improve among US investment grade issuers, based on corporate leverage statistics and cash balances. Gregory Peters, our chief credit strategist, has noted that 60% of issuers have reduced leverage fairly consistently over the past six quarters (see “A Fundamental Improvement,” March 22, 2004). Valuations, for the most part, already reflect this bondholder-friendly environment, although our credit analysts can point to at least some single-name opportunities. We have argued in previous research that structurally levering credits that are lightly levered themselves is a good way to express a strong fundamental credit view (see Chapter 74). Six months since our initial thoughts, credit fundamentals are even stronger, and there are numerous opportunities to express this view through specific first-to-default baskets. In particular, we find credits with little outstanding debt, large cash balances relative to outstanding debt, or debt payment schedules that are further off (say beyond five years) to be attractive candidates for inclusion in baskets. We go through several examples in this chapter. Liquidity, though, is always the number one question when investors consider customized structured credit opportunities. In the 18 months or so that we have closely tracked the first-to-default segment of the correlation markets, transparency has improved and, equally importantly, flows have rapidly increased, which we comment on in more detail, as well. SELLING FTD PROTECTION – INSURANCE COMPANIES ONLINE
The biggest theme in the first-to-default market is the entrance of insurance companies as sellers of protection. While this flow is not necessarily new and has certainly taken some time, we find quite a number of insurers (life insurers in particular) today who have investment programs in place. The motivation for getting involved has been the quintessential desire for yield, along with an appetite for credit risk in “non-traditional” names (highly rated issuers and non-US or off-the-run credits, where fundamentals are good, but spreads are tight). In fact, both of these sentiments have overshadowed the potential earnings volatility insurers take on for getting long credit risk through derivatives instead of cash instruments. We expect the insurance flow into the first-todefault market to be a continuing theme. Away from insurers, the market continues to broaden, but is still fragmented in our view. We see some selling of first-to-default protection in the hedge fund community, either outright or versus deltas. There are also pockets of interest in shorter maturities (less than five years) on potentially riskier names, a strategy that we support as well when credit fundamentals are strong.
518
BUYING FTD PROTECTION – FAR FROM COMPELLING?
In the current market environment, buying first-to-default protection outright has been far from a compelling trade, as there has not been nearly enough spread decompression (or idiosyncratic risk) to justify the levered premiums for these baskets. Yet we argue that buying first-to-default protection is a good hedge for those who take on significant idiosyncratic or concentration risk in their credit portfolios. The argument is that not all credits (where perceived credit risk is high) will ultimately default, but reducing concentration is important nevertheless. As such, carefully placing credits in a series of baskets and buying first-to-default protection will save premium. Bank loan portfolios are a natural application, and portfolio managers are increasingly buying first-to-default protection as an alternative to outright single-name protection positions. Yet the flows here are still relatively small and may not pick up substantially until banks’ appetite for protection picks up. Another very good use of first-to-default protection is in portfolios of equity and mezzanine tranches, given the significant jump-to-default risk. Those who manage portfolios of these risks spend a good amount of time hedging single-name risk for the obvious reasons, but the idea of spreading the idiosyncratic risk around in first-to-default baskets to save premium outlays is appealing, in our view, and will likely grow as liquidity grows in small baskets. THE UNLEVERED ALL-STARS
As we examine corporate leverage statistics for the generally healthy investment grade market, we find some extremely fundamentally fit corners of the market. In particular, from a default risk perspective, three themes strike us as interesting. •
Out of a 286 issuer universe in Gregory Peters’s study, 40 credits have Debt/EBITDA ratios that are below 1.0x (the investment grade median is 2.1x, the average is 2.5x).
•
The median ratio of cash to debt is 14%, but 27 credits have more cash than debt on balance sheets.
•
While the average credit has almost half of its outstanding public debt due in the next five years, 40 credits have less than 20% of their public debt due in the next five years. Twenty of these credits have less than 10% of debt due in the next five years. Thirty credits have virtually no debt due over the next three years.
While valuations, in many cases, reflect this fundamental fitness, medium term default risk is very low for these names, barring any major change in a company’s capital structure. For fundamentally minded investors, these are strong arguments for increasing exposure to default risk. One point we would like to drive home is the distribution of debt payments. In the default swap markets, it is possible to have a liquidity point for a credit even if the credit has little or no debt due on or before that date (e.g., five-year CDS for a credit with most of its debt due after five years). In the corporate bond market, on the other hand, this is impossible, by definition.
Please see additional important disclosures at the end of this report.
519
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 76
exhibit 1
The Healthy Tail of an Improving Leverage Distribution Debt/EBITDA (Gross/ Rent Adj.) 0.14
Cash/Debt 1,012%
% Public Debt Due Next 5 Years 40%
Home Depot Inc.
0.20
376%
100%
Exxon Mobil Corp.
0.25
115%
62%
3M Co.
0.61
63%
53%
ChevronTexaco Corp.
0.73
42%
72%
Apache Corp.
0.74
1%
13%
Credit Dell Inc.
Coca-Cola Co.
0.80
64%
69%
Newmont Mining Corp.
0.85
135%
55%
Applied Materials Inc.
0.89
978%
63%
Occidental Petroleum
1.10
14%
52%
Hewlett-Packard Co.
1.39
169%
84%
Wyeth
1.48
61%
15%
0.89/1.88
8%
10%
0.54/2.8
483%
0%
2.13
14%
42%
Kohl’s Corp. Limited Brands Investment Grade Median (286 Issuers)
Note: Data as of 12/03 or 9/03 when unavailable Source: Morgan Stanley, FactSet, Bloomberg
BASKET IDEAS
Levering the fundamentally fit is a good way to play a constructive view on default risk. In Exhibit 2 we show three baskets that specifically implement the fundamental themes we discussed above, although investors could certainly mix and match the themes in a basket context. The first basket plays on the low leverage theme, with five credits whose average Debt/EBITDA ratio is 0.95 (versus the 2.13 median for investment grade issuers). The seller of protection would earn a premium of 115 bp (or 79% of the total spread). The second basket carries the cash on balance sheet theme, with all issuers having cash levels on par with or significantly greater than debt. This earns a premium of 110 bp (80% of the total spreads). The last basket is perhaps the most extreme. Selling first-to-default protection on five technology companies that have little or no debt outstanding is really a play on the risk premium in the marketplace. This basket would pay 60 bp for very little actual default risk, which is made clear by examining the capital structures of these companies. The debt/equity ratios of Dell, Intel and Oracle are 0.57%, 0.65% and 0.51%, respectively, while Microsoft and Cisco have no outstanding debt. We find these ideas compelling from a default risk perspective, but caution that mark-tomarket risks, which can be driven by a variety of sources including increasing risk premium or M&A activity, should not be ignored. Selectively using single-name protection (for worrisome names) in basket strategies can help mitigate some of this risk.
520
exhibit 2 Ticker Low Leverage Basket
Levering the Fundamentally Fit – 3 FTD Basket Ideas Name
Spread (Mid)
APA
Apache Corp.
KSS
Kohl’s Corp.
39
OXY
Occidental Petroleum
36
AMAT
Applied Materials
26
DELL
Dell Inc.
19
Total Spread
25
145
FTD Spread (Bid/Offer)
115/126
FTD as % of Total
79%/87%
Cash Rich Basket HPQ
Hewlett-Packard Co.
LTD
Limited Brands
31 32
NEM
Newmont Mining
33
HD
Home Depot Inc.
15
AMAT
Applied Materials
26
Total Spread
137
FTD Spread (Bid/Offer)
110/120
FTD as % of Total
80%/88%
Little or No Debt Basket DELL
Dell Inc.
INTC
Intel
17
MSFT
Microsoft
13
CSCO
Cisco
15
ORCL
Oracle
15
Total Spread FTD Spread (Bid/Offer) FTD as % of Total
19
79 60/65 76%/82%
Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
521
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 77
Basket Weaving 101
August 13, 2004
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Primary Analyst: Anisha Ambardar Angira Apte The apparent difference in relative pricing between large baskets and small baskets has been a subject we have continually had interest in as the structured credit market has developed. In the early days, we pointed to the difference in the absolute level of correlation as an indication of the high level of idiosyncratic risk associated with first to default positions as compared to the much broader credit exposure in large baskets (see Chapter 9). Today, we have the opportunity to look at how these relationships have developed over time. We also note that the recent introduction of first-to-default baskets created from the names in the CDX index has the potential to bring the large and small basket markets closer together, creating opportunities to trade the relative value between the two families of instruments. In this chapter, we examine the interrelationships among benchmark products in both large and small basket spaces during this year. We then attempt to use historical pricing from actively traded benchmark products to measure the implied performance of large and small basket products based on the CDX index. exhibit 1
Implied Correlations Remain at Different Levels
23%
49%
47% 21%
45%
43% 19% 41%
39%
37%
35% Oct-03 Nov-03 Dec-03 Jan-04
Source: Morgan Stanley
522
Average FTD Correlation (Left Axis)
17%
0-3% TRACX Correlation (Right Axis)
15% Feb-04 Mar-04 Apr-04 May-04 Jun-04
Jul-04
A BRIEF “HISTORY” OF BASKETS
Exhibit 1 illustrates the implied correlations for the 0-3% tranche of the Dow Jones TRAC-X Series 2, as well as the average correlation of our six benchmark FTD baskets since October 2003. We can see that the implied correlations continue to be at different levels on an absolute basis but did move in reasonably similar patterns in late 2003 through early 2004. Recently, there has been somewhat of a divergence, as FTD correlations have remained relatively flat while the 0-3% tranche implied correlation has increased. exhibit 2
Large and Small Basket Strategies Underlying
Strategy 1 0-3%
CDX
Strategy 2 Basics CDX FTD CP & Retail CDX FTD Energy CDX FTD Financials CDX FTD TMT CDX FTD
Notional Exposure Per Credit Total Portfolio Exposure Position Notional Number of Credits
F, BOMB, DPH, DOW, IP MO, ABS, SRAC, SWY, TSN DUK, HAL, FE, D, COP GECC, CIT, CFC, AIG, COF T, TWX, EDS, MOT, VZ Strategy 1 3,333,333
Strategy 2 10,000,000
416,666,667
250,000,000
10,000,000
50,000,000
125
25
Source: Morgan Stanley
Given the differences in structure between the large and small baskets, implied correlations alone cannot tell the full story. With this in mind, we thought it would be a worthwhile exercise to examine how the strategies of going long credit through a 0-3% tranche of CDX while shorting a portfolio of CDX FTD baskets would have performed. In Exhibit 2, we describe two strategies we studied historically. The first is a position in the 0-3% tranche of CDX and the second is a portfolio of five CDX FTD baskets, which is weighted to be roughly equivalent on a premium basis. As we currently have limited actual price history for both CDX tranches and CDX first-to-default baskets, we recreated theoretical values for the baskets from singlename spread levels for purposes of this study. We generated the theoretical values using the par spreads for all the underlying credits and the implied correlations from similar, but actively traded benchmark products. In the case of the CDX FTD baskets, we used our benchmark FTD basket implied correlations and in the case of tranched CDX, we used the implied correlations from the same tranches of Dow Jones TRAC-X Series 2.
Please see additional important disclosures at the end of this report.
523
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 77
We then modified the initial notional amounts to generate a combined portfolio that would minimize the standard deviation of the total package profit/loss over the period. The result was to increase the notional of the CDX tranche by approximately 38% to $13.8 million. Exhibit 3 shows the implied profit/loss for both of the strategies and the combined package over the historical period. exhibit 3
Theoretical P/L of Hedged Position
2,000,000
Short CDX FTD Portfolio Long 0-3% CDX IG NA Hedged Position
1,000,000
0
-1,000,000
-2,000,000 Jan-04
Feb-04
Mar-04
Apr-04
May-04
Jun-04
Jul-04
Source: Morgan Stanley
DIFFERENT RISKS ACROSS DIMENSIONS
While we would expect some level of consistency between these two strategies (both being first loss positions in overlapping portfolios), there are several factors that can affect returns in a different manner than our historical analysis indicates. First, the breadth of risk is much greater for the tranched CDX position with its 125name underlying portfolio. The first-to-default exposure is much more idiosyncratic in nature, with only one-fifth the number of names and significantly greater notional exposure per name (see Exhibit 2). Additionally, the aggregate notional amount is much greater for the FTD portfolio (both the initial and adjusted portfolios). Another way to measure this difference in broad-based versus idiosyncratic exposure is to consider the sensitivity of the strategies to individual credit spread moves and marketwide spread moves, which we summarize in Exhibit 4. Both strategies have similar sensitivities to the moves of any single credit within the portfolio but the CDX FTD strategy is much less sensitive to moves in all the names in the underlying portfolio.
524
exhibit 4
DV01 Tightest Credit (bp) DV01 Widest Credit (bp)
FTDs vs. Tranches – Spread Sensitivity Varies Strategy Risk Metrics (in BP) 0-3% of CDX CDX FTD Portfolio 0.53 0.42 0.76
0.64
DV01 All Credits (bp)
78.26
12.96
Correlation DV01 (bp)
74.93
6.92
Source: Morgan Stanley
Another fundamental difference is the sensitivity of the positions to changes in correlation. Again in Exhibit 4, we summarize the correlation sensitivity of the two strategies. Because the 0-3% tranche is a “thinner” exposure to a broader universe than the FTD basket portfolio, the sensitivity to correlation changes is much more significant, reflecting the more levered nature of the 0-3% tranche. BASKET WEAVING
To date, the markets for large and small baskets appear to have been trading relatively independent of one another, which is a dynamic we find interesting. With the increased liquidity in both the CDX FTD baskets and tranches, we would keep an eye out for opportunities to trade relative value using combinations of large and small baskets.
Please see additional important disclosures at the end of this report.
525
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 78
When Mergers De-Lever
January 7, 2005
Primary Analyst: Sivan Mahadevan Primary Analyst: Peter Polanskyj Ajit Kumar, CFA One theme that appears to be a common sentiment in several sectors this year is the potential for M&A activity. The markets have experienced a decent amount of it already in the telecom, media, and retail sectors, and our sector analysts feel that there is potentially more to come in the same sectors as companies look for strategic growth opportunities, consolidation benefits or, simply, synergies. One can build many investment strategies centered on potential changes in a company’s capital structure due to mergers, with long/short pairs or debt/equity strategies being the most natural. In the case of M&A activity, the key driver of performance is getting the market direction right on the various legs, but that has proven frustrating, as debt holder-unfriendly activity has not necessarily been treated as bad news by credit market participants, at least not recently. For investors with first-to-default basket-type exposure to a portfolio of credits where there is consolidation, a merger represents a reduction in risk, as the number of potential reference entities that can default is reduced (by one) when two companies merge. From a default correlation perspective, one can think of this as rising correlation because the correlation of the two merged companies goes to 100%. The seller of protection would benefit from this action alone, although a widening of spreads on the underlying names could offset the benefit if the names were not deltahedged or they underperformed their deltas. This structural consolidation idea is not new, and we addressed it initially when there was a mini-wave of M&A activity at the end of 2003 (see Chapter 75 of this book and Chapter 23 of Credit Derivatives Insights – Single Name Instruments & Strategies, 2nd Edition, February 2006). We find the current landscape, particularly in the retail, media, and telecom sectors (across both high yield and investment grade) to be potentially fruitful for such ideas as well, given our analysts’ views. We caution, though, that getting the direction of spreads right is still important, which can be based on anticipated ratings actions, as well as impact of business sizes. Furthermore, a key detail to focus on is the use of “replacement” language in standard FTD contracts, which can take away any consolidation advantage and introduce other (unwanted) risks.
526
REPLACEMENT LANGUAGE – AN IMPORTANT DETAIL
As an aside, albeit an important one, there are two ways one can trade an FTD basket in the market today: with or without merger replacement language. Replacement language refers to the right the buyer of protection has to replace a reference entity that merges with another, when both appear in the basket. The terms of the replacement (when replacement language is selected) are complicated, allowing the buyer of protection to select at least three reference entities that meet certain guidelines including sector, spread, and geographic region (all similar to the merged entity). The seller of protection must then select one of these reference entities, and it would then be used as the “replacement” credit in the basket for the merged entity. Replacement language creates a risk-management headache because there can be a lot of subjectivity associated with selecting the replacement entity, even with the marketstandard guidelines. The seller and buyer of protection will have opposite views on correlation, for example, and the delta hedging shifts can be onerous. Furthermore, we are not aware of any good tests that this replacement language has been through yet, and we are certain that such tests would highlight other shortcomings, much like the market saw with the original restructuring credit events years ago. For these reasons, we continue to recommend that investors consider FTD baskets without merger replacement language (see Chapter 75 for our initial thoughts). As such, the seller of first-to-default protection is effectively long a merger option, as two entities that merge would result in a “structural de-levering” of the basket, which benefits the protection seller, and is a key motivation of our FTD basket ideas. WIRELINE WARS
The battle between long distance and local carriers continues today and our telecom research team sees consolidation across these business lines as a reasonable outcome given the evolving economics of the telecom space. Given our analysts’ view that these mergers would likely not be significantly negative credit events for the acquirer in either case, we find the strategy of selling protection in a first-to-default basket (without replacement language) to be an attractive way to play the current dynamics in the industry. Exhibit 1 illustrates a basket containing the major RBOCs and likely long distance players involved in this type of merger activity. The basket would benefit from structural de-levering driven by a merger between any of these players but would have the most upside in the acquisition of a long distance carrier by an RBOC. As with the wireless basket we recommended in earlier research, there is also the potential that multiple mergers occur within the basket (see Chapter 23 of Credit Derivatives Insights – Single Name Instruments & Strategies, 2nd Edition, February 2006).
Please see additional important disclosures at the end of this report.
527
Structured Credit Insights – Instruments, Valuation and Strategies
chapter 78
exhibit 1
Positioning for Wireline Consolidation
Name Wireline FTD Basket
Spread Mid 272/318
AT&T
111
MCI Inc
230
Bell South
23
SBC Communications
24
Verizon Global
21
Total Spread
409
% of Total 67%/78%
Source: Morgan Stanley
RETAIL MERGERS WOULD BE A RE-LEVERING EVENT
The retail sector provides another perspective on merger activity, as increasingly acquisitive behavior would likely need to be financed in the credit markets. As such, structural de-levering gains in an FTD basket due to mergers would likely be offset (or more than offset) by issuance-driven spread widening. The recent acquisition of Marshall Fields by May Department Stores provides an interesting example. In order to finance the merger May issued $2.2 billion of bonds and its default swap premium widened from 48 bp in early May 2004 to 76 bp in early July 2004. We created a long/short package (see Exhibit 2) made up of a long position in Dillard’s and a short position in Federated which has four times the notional. This package is positive carry and would benefit if Federated was acquisitive (in a credit-unfriendly manner) whether or not the target was Dillard’s. The package would also benefit if Dillard’s was acquired by any entity in a credit-enhancing transaction. We have increased the short notional amount because it creates a package that is net short. exhibit 2
Federated Dillard's Package Source: Morgan Stanley
528
Retailer Merger Long/Short Package
5 Year CDS Bid 30 185
Offer 35 205
Notional (40,000,000)
Net Carry (on Long Notional)
10,000,000 (30,000,000)
45.0
MEDIA INTEGRATION
In the media space, we have designed a basket (see Exhibit 3) that would benefit from consolidation, integrating content providers with the owners of distribution channels. Most of the potential acquirers trade fairly tight while the potential targets/merger partners are at wider levels, making the possible upside from structural de-levering that much greater, assuming the acquirer’s spreads do not markedly widen. The risk that the acquirers widen in the merger should not be ignored, however, and can be hedged to some degree by delta hedging these credits while remaining unhedged for the likely targets. In the case of the media basket, this package would still be a positive carry trade. exhibit 3 Name Media FTD Basket Viacom
Media Content & Distribution Merger Basket Spread Mid 272/321 27
Cablevision
190
Echostar
152
Time Warner Disney Total Spread
% of Total 62%/74%
38 29 436
Source: Morgan Stanley
Please see additional important disclosures at the end of this report.
529
Section H
Glossary
Please see additional important disclosures at the end of this report.
Structured Credit Insights – Instruments, Valuation and Strategies
Glossary ARBITRAGE CDO
CDO designed to take advantage of the funding gap (i.e., earn the spread between the yield of the underlying assets and the cost of CDO liabilities). ASSET BACKED COMMERCIAL PAPER (ABCP)
An ABCP program is composed of a bankruptcy-remote special purpose vehicle (SPV), or conduit, that issues commercial paper and uses the proceeds of such issuance to obtain interests in various types of assets, either through asset purchase or secured lending transactions. ATTACHMENT POINT
Defines the amount of losses in the reference pool of assets that needs to occur before the tranche starts to experience losses. Can be expressed as a percentage or as an absolute value. AVAILABLE FUNDS CAP
The interest on US home equity loan securitizations is capped at a pre-specified rate called the available funds cap rate. This feature introduces interest rate risk to such securitizations and credit default swaps where the reference obligation is a home equity loan. BACK-END DEFAULT RISK
The default risk of a portfolio that is weighted towards the maturity date of the tranche, rather than in the near future. Opposite of Near-Term Default Risk. BALANCE SHEET CDO
CDO created to allow the sponsor (typically a bank) to remove credit exposure from its balance sheet in order to achieve regulatory capital relief. Balance sheet CDOs are typically static. BASE CORRELATION
Base correlation methodology considers all non-equity tranches as a portfolio of long 0% to the detachment point tranche and short 0% to the attachment point tranche. The implied correlation figures calculated in this fashion from given tranche prices are referred to as base correlation. BERMUDA-STYLE OPTION
An option that can be exercised by the holder only on a series of preset dates up to the final maturity date. Bermuda-style call options are typically embedded in senior CDO notes and held by equity holders, thus providing equity holders the opportunity to unwind a CDO structure should deleveraging occur.
532
BERNOULLI DISTRIBUTION
A probability distribution of an event with two possible outcomes. Defaults can be represented by a Bernoulli distribution, where p equals the probability of default and 1p equals the probability of a no-default scenario. BESPOKE
Typically refers to a customized portfolio/collateral pool (may be customized by either the buyer or seller). BINOMIAL EXPANSION TECHNIQUE (BET)
Valuation methodology used by Moody’s to rate cash CDOs. Reduces a given portfolio to a number of uncorrelated assets (see Diversity Score) and calculates the expected loss on each of the tranches by varying the number and probability of defaults. The expected loss values are then used to assign ratings to the various tranches. CARRY
The differential between the yield on an investment and the cost of funding the position. Also the difference between buy and sell legs in a relative value relationship. Typically measured in basis points. CASH CDO
A stand-alone special purpose vehicle that invests in a diversified pool of assets in a funded fashion, as opposed to using credit derivatives (unfunded). Investments are funded through the issuance of multiple classes of securities. CASH FLOW CDO
A CDO whereby the principal and interest of the liabilities of the structure are paid using the cash flows generated by the underlying collateral. CASH FLOW METHODOLOGY
Values a CDO by discounting the scheduled cash flows of the various tranches over one or multiple paths. Also incorporates default, recovery and reinvestment assumptions as potential impacts on the expected cash flows. CASH FLOW TRIGGERS
Structural features of a cash CDO that can determine and change payouts to different tranches based on the performance of the underlying collateral. CDO EVALUATOR
Proprietary S&P model used to rate various CDO tranches. Uses a Monte Carlo simulation to generate a probability distribution of defaults for each asset, taking individual credit ratings, number of assets, industry concentration and default correlation into consideration. CDO SQUARED
A CDO whose underlying collateral pool includes tranches of other CDO structures.
Please see additional important disclosures at the end of this report.
533
Structured Credit Insights – Instruments, Valuation and Strategies
COLLATERAL MANAGER
Party responsible for making the investment and trading decisions in the reference portfolio. Decisions are typically made within pre-designated guidelines. COLLATERALIZED BOND OBLIGATIONS (CBO)
A securitized pool of assets comprised predominantly of bonds. Examples of collateral may include investment grade bonds, high yield bonds, convertible debt, mezzanine debt or emerging market debt. COLLATERALIZED LOAN OBLIGATIONS (CLO)
A securitized pool of assets comprised predominantly of loans, which may be investment grade, high yield or both. COLLATERALIZED MORTGAGE OBLIGATIONS (CMO)
A securitized pool of assets comprised predominantly of mortgages. COMPOUND CORRELATION
The implied correlation of a tranche, given its market price. Certain tranches may have more than one solution for a given price. CONVEXITY
The changes in delta given a change in underlying spreads. Typically measured as a ratio of the mark-to-market of a tranche, given a move of 100 bp in underlying spreads to the PV01 of a tranche multiplied by a factor of 100. COPULA FUNCTION
Function used to transform a given set of loss distributions for individual credits into a joint loss distribution for a portfolio of credits. CORRELATION SENSITIVITY
See RHO. CORRELATION SKEW
When different tranches referencing the same underlying portfolio trade at different implied correlations. Also refers to the difference in correlation for attachment and detachment points of a tranche. CREDIT-LINKED NOTE
When credit risk is embedded into a note that pays a fixed or floating coupon reflecting the riskiness of a credit. See Funded Form. CREDIT ENHANCEMENT
Feature of a CDO designed to decrease the credit risk of the structure. Forms of credit enhancement may include overcollateralization, guarantees from insurers (see Wrapped Tranches) and letters of credit.
534
CROSS SUBORDINATION
Provision of a CDO-Squared that allows the unused subordination of an inner tranche to be transferred to another inner tranche for the purpose of absorbing losses from the underlying inner CDO. DEFAULT-ADJUSTED CASH FLOWS
The expected cash flows of a CDO that are modified to reflect the likelihood of default. Multiply the probability of default by the expected cash flow and discount using LIBOR to derive a present value. DEFAULT PV
The present value of all expected default losses to the maturity of a credit derivative contract. DELEVERING
In a CDO structure, paying down of liabilities with excess spread. DELTA
Tranche sensitivity to changes in underlying portfolio spread, measured as a ratio of tranche PV01 to portfolio PV01. The delta of a tranche generally increases as the subordination of the tranche decreases. DELTA MIGRATION
Refers to the instability of tranche deltas with respect to movements in index spreads, changes in correlation and the passage of time. DELTA-NEUTRAL
When the value of a tranche is unaffected by small changes on underlying spreads in the underlying index because of an offsetting hedge using the index or other tranche. DETACHMENT POINT
Defines the amount of losses in the reference pool of assets that need to occur for a complete loss of principal for the tranche. Can be expressed as a percentage or as an absolute value. DIVERSITY SCORE (N)
Moody’s-developed numerical statistic (N) reflecting the number of uncorrelated homogenous assets with identical default probabilities and equal par values, reduced from a given portfolio. DOUBLE BINOMIAL METHOD
Used to value CDOs comprised of two distinct pools of underlying assets. Each pool is assumed to be uncorrelated and each has a distinct default probability and diversity score.
Please see additional important disclosures at the end of this report.
535
Structured Credit Insights – Instruments, Valuation and Strategies
DOW JONES CDX NA HY INDEX
The synthetic index of 100 non investment grade entities domiciled in North America. Composition of the index is rules based and determined by a dealer liquidity poll. CDX HY index is rolled over every six months in March and September. DOW JONES CDX NA IG INDEX
The synthetic index of 125 investment grade credits in North America, as determined by a consortium of dealers. Rolls over on September 20 and March 20 of each calendar year. ENHANCED EQUIPMENT TRUST CERTIFICATE (EETC)
Tranches of securitized debt collateralized by specific assets and equipment of a firm, with a waterfall structure in case of bankruptcy. Typically issued by industrial companies and airlines. EQUITY
The most subordinated tranche of a CDO. Also known as a first-loss piece, as any losses experienced by the underlying collateral pool will reduce the notional of the tranche before flowing upwards to the other tranches. Typically trades on a points upfront basis for synthetic indices. FIRST-LOSS PIECE
See Equity. FIRST-TO-DEFAULT BASKET
A levered investment on the default risk of a reference portfolio where the buyer of protection is protected against the first default of the basket, at which point the contract is settled and terminated. Basket size typically ranges from four to ten credits and can be funded or unfunded. FUNDED FORM
When credit risk is embedded into a note that pays a fixed or floating coupon reflecting the riskiness of a credit and the protection seller pays the full principal amount at the inception of the trade. If a credit event occurs, losses result in a writedown of the principal; otherwise known as a Credit-Linked Note. FUNDING GAP
The difference between the yield of the assets in the reference pool and the yield of the CDO's liabilities. HAZARD RATE
The forward probability of default over a specified time horizon. Can be inferred from CDS premiums or asset swap spreads. HG ABS CDO
CDO transactions with average underlying portfolio credit quality above A3/A- at deal inception.
536
HYBRID SETTLEMENT
The process by which all standard indices and index tranches settle in the event of a default. Under hybrid settlement, the buyer of protection identifies the amount of bonds they intend to physically deliver to the seller, who then conducts an auction for the bonds. The price of the auction is used to compute losses if applicable and the amount of subordination loss. I-GAMMA
Sensitivity of a tranche value to jump-to-default risk or changes in the spread distribution of the underlying portfolio. IDIOSYNCRATIC RISK
Risk that is specific to a security or issuer/company and unrelated to market risk. Such firm-specific risk can be diversified away. Opposite of Systemic Risk. IN-THE-MONEY
Refers to tranches where default losses are expected to reduce the notional as implied by the index premium. Such tranches are typically traded on a points upfront basis. INNER CDO
Refers to the CDO tranches comprising the reference portfolios in a CDO-Squared. INTEREST COVERAGE TEST
Ratio of the expected interest generated by the underlying collateral pool to the interest due on the given tranche, plus the interest due on all tranches senior to it. ITRAXX EUROPE
The synthetic index that consists of 125 equally weighted credit default swaps on European entities. Composition of the index is rules based and determined by a dealer liquidity poll. iTraxx Europe is rolled over every six months in March and September. JUMP-TO-DEFAULT RISK
The risk of a credit spread gapping significantly wider to imply a high probability of default in the near term, as opposed to a gradual spread widening. JUNIOR TRANCHE
Tranche that is subordinated to the senior tranche in terms of its claim on the underlying collateral pool. Ratings and spreads on junior tranches will reflect the increased credit risk of the securities. Junior tranche investors are typically long correlation. LEVERAGE
With respect to CDOs and other securitized assets, leverage refers to the size of the equity tranche relative to the total size of the structure.
Please see additional important disclosures at the end of this report.
537
Structured Credit Insights – Instruments, Valuation and Strategies
LSTA
Loan Syndication and Trading Association. Formed by international financial institutions, LSTA aims to develop standard settlement and operational procedures, market practices and other mechanisms to more efficiently trade par and distressed bank debt. MAKE-WHOLE PREMIUMS
When a security or debt tranche is called by either the issuer or the equity tranche holders, a premium payment (in addition to the par payment) may be required to compensate the debt holders for the early retirement of debt. MANAGED
When the reference portfolio may be traded by the collateral manager according to designated guidelines and restrictions. MARK-TO-MARKET
The current market value of a security or a position that reflects any unrealized gains or losses since inception. MARKET VALUE CDO
Cash CDO in which the principal and interest of the structure are paid using the proceeds generated by the sale and trading of the underlying collateral. MARKET VALUE METHODOLOGY
Values a CDO by equating the market value of the assets (the underlying collateral) to the market value of the liabilities (comprised of the debt equity and management fees). MEZZ ABS CDO
CDO transactions with average underlying portfolio credit quality of A3/A- or below at deal inception. MEZZANINE TRANCHE
A Junior Tranche that has some subordination. M-GAMMA
Sensitivity of a default-neutral position to parallel shifts in spreads of underlying credits. MODIFIED MODIFIED RESTRUCTURING (MOD-MOD-R)
Under this definition, the main difference from Mod-R is that the protection buyer can deliver a deliverable obligation with maturity up to 60 months after restructuring (in the case of the restructured bond or loan) and 30 months in the case of all other deliverable obligations. The majority of European-domiciled credit default swaps trade on a Mod-Mod-R basis.
538
MODIFIED RESTRUCTURING (MOD-R)
In the case of a restructuring credit event, the protection buyer must deliver obligations with a maturity date prior to a) 30 months following the restructuring and b) the latest final maturity date of any restructured bond or loan, but not shorter than the CDS contract. Most North American-domiciled credit default swaps trade on a Mod-R basis. MONTE CARLO SIMULATION
Technique that uses randomly generated values as inputs for given variables to construct a probability distribution of the resulting outcomes. MULTISECTOR CDO
CDO whose underlying collateral pool references a combination of bonds and loans from different credit market sectors (e.g., corporate credits, ABS, MBS, CMBS, Emerging Markets, etc.). MULTIVARIATE GAUSSIAN COPULA
The correlated multivariate normal distribution used to generate a portfolio loss distribution given single-name loss distributions. NEAR-TERM DEFAULT RISK
Risk that the underlying portfolio/tranche is expected experience defaults in the near future. Typical of equity tranches. Opposite of Back-End Default Risk. NO RESTRUCTURING (NO-R)
Provision within a CDS contract which does not consider restructuring to be a valid credit event. No-R protection typically trades cheaper than Mod-R protection. NTH-TO-DEFAULT BASKET
Similar to a First-to-Default basket, it is a levered investment on the default risk of a reference portfolio except it is the Nth credit event that triggers the contract. May be funded or unfunded. OUT-OF-THE-MONEY
Refers to tranches where the likelihood of default losses penetrating the tranche notional is low. Protection purchased on senior tranches is typically out-of-the-money. OUTER CDO
Refers to the CDO whose underlying portfolio references other CDO tranches in a CDO-Squared. OVERCOLLATERALIZATION TEST
Valuation metric comparing the value of a CDO's underlying collateral to the structure's liabilities. A ratio greater than 1 represents an over-collateralization of the CDO's liabilities. See Par Coverage Test.
Please see additional important disclosures at the end of this report.
539
Structured Credit Insights – Instruments, Valuation and Strategies
OVERLAP RATIO
Ratio of bonds overlapping between a CDO's collateral portfolio and a standardized synthetic index's reference portfolio, assuming the CDO investor uses that synthetic index as a hedge. PAR COVERAGE TEST
An overcollateralization test used to measure the overcollateralization of senior and junior tranches. To calculate, divide the sum of the par value of performing assets and the expected recovery value of defaulted assets by the par amount of the tranche plus the par amount of any tranche senior to it. PARI PASSU
Latin term meaning at an equal pace or without partiality. In the event of a liquidation, creditors that rank pari passu would have equal entitlement to the assets and hence would be paid pro rata in accordance with the amount of their claim. PAYMENT-IN-KIND (PIK)
Provision where the interest on a tranche/security is deferred and instead added to the principal amount of those tranches in order to mitigate a potential default by the tranche and extend the life of the structure. PREMIUM PV
The present value of all expected premiums to the maturity of the CDS contract. PREPAYMENT
When an issuer partially or completely pays down the debt ahead of the scheduled maturity date, thus shortening the weighted average maturity of the debt. PRO-RATA PAYDOWN
A structural feature (found most commonly in high-grade ABS CDOs) that allows all classes of debt in the capital structure to be paid down in proportion to their outstanding balances as long as there has not been any breach of a coverage test and a certain proportion of the collateral balance is still outstanding. PV01
The change in the price of a tranche given a 1 bp change in each of the underlying credits in the portfolio. RAMP-UP PERIOD
Period of time where the CDO collateral manager invests in assets using the proceeds from the issuance of debt tranches from the CDO. RECOVERY RATE
The value of the deliverable obligation received by the protection seller when a credit event occurs, calculated as a percentage of par. Recovery rates tend to be lower in a high default rate environment.
540
REINVESTMENT PERIOD
Length of time over which the CDO manager may reinvest the proceeds of matured or traded securities. When this period ends, the underlying portfolio will become a static one. REPLACEMENT LANGUAGE
Provision within a Nth-to-default basket where the buyer of protection has the right to replace a reference entity that merges with another entity in the basket. Replacement entities must typically meet certain guidelines, such as sector and spread. RE-RATING METHODOLOGY
Values a CDO by inferring a rating for a CDO tranche by calculating its expected loss and considering its risk characteristics. The tranche is then compared to similar structures in the market to determine a comparable spread. RISK-NEUTRAL DEFAULT PROBABILITY
Probability of default implied by CDS pricing for a given time period. A risk-neutral default probability results in expected losses that match the present value of the CDS premium. RHO
Change in tranche value due to changes in correlation. SAMPLING RISK
The risk that a sample does not reflect the true characteristics of the population. For example, a credit portfolio, while having the same expected losses as the entire universe of credits, might have a more fat-tailed distribution. SELF-MANAGED SYNTHETICS
A structure that allows investors to take exposure in tranche form to a portfolio that he or she manages. Typically, no other exposure in this portfolio would be distributed by the dealer. SENIOR TRANCHE
Tranche bearing the least credit risk in the structure as it has a priority claim on all cash flows generated by the collateral pool. All other tranches are subordinated and their principal remains outstanding until the senior tranche is fully paid off. Generally carries the highest rating in the structure. In synthetic indices, senior tranche investors are typically short correlation. SHARPE RATIO
Excess return on an investment (i.e., return over the risk-free rate) per unit of risk (as measured by standard deviation of the returns). SINGLE TRANCHE SYNTHETIC CDO
Form of arbitrage CDO that allows the investor to customize the reference pool of assets by selecting the credits, as well as the attachment and detachment points.
Please see additional important disclosures at the end of this report.
541
Structured Credit Insights – Instruments, Valuation and Strategies
SPECIAL PURPOSE VEHICLE (SPV)
A stand-alone, bankruptcy-remote entity whose purpose is limited to the acquisition and financing of specific assets. Cash and synthetic CDOs, as well as other securitized asset pools, are typically issued from an SPV. STATIC
When the reference portfolio is fixed at inception. STRUCTURAL DELEVERING
Refers to the decrease in overall leverage when two entities in a Nth-to-Default basket (without replacement language) merge. STRUCTURED CREDIT BID
The proliferation of investors (protection sellers) in synthetic CDOs, resulting in a significant tightening of credit spreads as dealers buy protection from these investors. SUB PRIME ABS
This refers to securitizations that have an underlying collateral pool of consumer credit with an average FICO score below 660. FICO is a credit score scale that uses a riskbased system to determine the possibility that the borrower may default on financial obligations to the lender. SUM OF SPREADS
The sum of the premiums on the underlying credits in a basket. A buyer of protection on an Nth-to-Default basket will pay a percentage of that sum of spreads, based on the default correlation assumption. SYNTHETIC RATED OVERCOLLATERALIZATION (SROC)
A measure of the sensitivity of a tranche in a synthetic CDO to future rating action, expressed as a ratio. 100% SROC implies an exactly sufficient credit enhancement to maintain rating on a tranche. Very loosely, SROC can be defined as (1 - Scenario Credit Loss)/(1 - Credit Enhancement). TERM LOAN
A term loan is a funded loan type wherein once borrowed funds are repaid, the funds cannot be borrowed again. Institutional term loans are usually taken out by leveraged borrowers (i.e., non-investment grade borrowers with debt/EBITDA greater than 2.0x) and repaid at maturity (there might be some minimal, back-ended principal repayments). They are normally longer dated compared to amortizing term loans (five to seven years) and may be prepaid at any time at par. Multiple tranches with varying maturities can co-exist within a facility. THETA
Change in tranche value due to the passage of time. TIME DECAY
Refer to Theta.
542
TRANCHE
A defined portion of credit risk within a securitized pool of assets. Each tranche has a specific attachment and detachment point, representing its exposure to losses that may reduce its principal amount. The priority of a tranche’s claims on underlying collateral varies, depending on whether it is equity, mezz, or senior within the capital structure. TRANCHE THICKNESS
The difference between the attachment and detachment point for a given tranche. Defines the maximum amount of losses that the tranche can experience. TRUPS CDO
Cash CDO transactions backed by trust preferred securities. TRUPS are subordinate to all debt but are senior to preferred and common stock, pay quarterly or semiannual interest, and are usually longer term. TRUPS are an important source of capital for smaller banks and insurance companies with limited access to capital markets. TRUSTEE
Party responsible for ensuring compliance with pre-designated trading and structural guidelines within a CDO structure. Periodically provides reports to investors detailing the status of the structure’s assets and liabilities, as well as compliance updates. UNFUNDED FORM
When no upfront payment is made by the seller of protection at the inception of the trade and no principal payment is paid at maturity. Payments between the buyer and seller of protection are comprised of premium payments and a termination payment should a credit event occur. Typical form of a plain vanilla credit default swap. UPFRONT PREMIUM PAYMENT
A payment made by the buyer of protection to the seller at the time of purchase. Certain tranches (e.g., equity and junior mezzanine) and credits with a high default probability tend to trade on a points-upfront basis, with or without a running premium. UPGRADE-DOWNGRADE RATIO
Compares the number of ratings upgrades versus the number of downgrades in a class of CDO tranches to determine the direction and magnitude of ratings migration. VECTOR
Approach used by Fitch to rate CDOs by assigning probabilities to expected losses. It uses a Monte Carlo simulation to model default distributions using individual asset default rates and asset correlations. VINTAGE
The year that the CDO was originally issued. WATERFALL MECHANISM
Typical structure of a securitized pool of assets with tranches of varying seniority, whereby losses are absorbed by the most junior tranches, while claims on the underlying assets are prioritized by seniority.
Please see additional important disclosures at the end of this report.
543
Structured Credit Insights – Instruments, Valuation and Strategies
WEIGHTED AVERAGE LIFE (WAL)
The weighted average period of time over which the principal of a tranche is expected to remain outstanding since the time of issuance. WEIGHTED AVERAGE MATURITY (WAM)
The weighted average period of time until the final principal payment of each tranche, weighted by the size of each tranche. WEIGHTED AVERAGE RATING FACTOR (WARF)
A numerical metric indicating the credit quality of a portfolio (which can vary across ratings agencies). Typically, the higher the WARF, the lower the credit quality of the portfolio. Most CDO structures will designate a maximum WARF value as guidelines. WEIGHTED AVERAGE SPREAD (WAS)
The weighted average spread of the underlying collateral/portfolio in a CDO structure. WRAPPED TRANCHES
Tranches that are guaranteed by bond insurers and are thus assigned ratings reflecting the creditworthiness of the bond insurer.
544
Disclosures Credit Products Rating Distribution Table (as of April 1, 2006) Coverage Universe Rating
Count
% of Total
Investment Banking Clients (IBC) Count
% of Total IBC
% of Rating Category
Overweight 85 37% 47 37% 55% Equal-weight 110 48% 65 51% 59% Underweight 32 14% 16 13% 50% 227 128 Total Coverage includes all companies that we currently rate. Investment Banking Clients are companies from whom Morgan Stanley or an affiliate received investment banking compensation in the last 12 months.
Analyst Ratings Definitions Overweight (O) Over the next 6 months, the fixed income instrument’s total return is expected to exceed the average total return of the relevant benchmark, as described in this report, on a risk adjusted basis. Equal-weight (E) Over the next 6 months, the fixed income instrument’s total return is expected to be in line with the average total return of the relevant benchmark, as described in this report, on a risk adjusted basis. Underweight (U) Over the next 6 months, the fixed income instrument’s total return is expected to be below the average total return of the relevant benchmark, as described in this report, on a risk adjusted basis. More volatile (V) The analyst anticipates that this fixed income instrument is likely to experience significant price or spread volatility in the short term.
IMPORTANT DISCLOSURES ON SUBJECT COMPANIES The information and opinions in this report were prepared by Morgan Stanley & Co. Incorporated and/or one or more of its affiliates (collectively, “Morgan Stanley”) and the research analyst(s) named on page one of this report. As of March 31, 2006, Morgan Stanley held a net long or short position of US$1 million or more of the debt securities of the following issuers covered in this report (including where guarantor of the securities): Saks Incorporated, Saks Incorporated, Nova Chemicals Corporation, Nova Chemicals Corporation, Dillard's, Inc., Federated Department Stores, Inc.. As of March 31, 2006, Morgan Stanley beneficially owned 1% or more of a class of common equity securities/instruments of the following companies covered in this report: Saks Incorporated Co. An employee or director of Morgan Stanley serves as an officer, director or advisory board member of Federated Department Stores. Within the last 12 months, Morgan Stanley has received compensation for investment banking services from Nova Chemicals Corporation, Texas Genco Holdings Inc. (Houston). In the next 3 months, Morgan Stanley expects to receive or intends to seek compensation for investment banking services from Saks Incorporated, Nova Chemicals Corporation, Texas Genco Holdings Inc. (Houston), Dillard's Inc., Federated Department Stores. Morgan Stanley policy prohibits research analysts from investing in securities/instruments in their MSCI sub industry. Analysts may nevertheless own such securities/instruments to the extent acquired under a prior policy or in a merger, fund distribution or other involuntary acquisition. Morgan Stanley is involved in many businesses that may relate to companies or instruments mentioned in this report. These businesses include market making, providing liquidity and specialized trading, risk arbitrage and other proprietary trading, fund management, investment services and investment banking. Morgan Stanley trades as principal in the securities/instruments (or related derivatives) that are the subject of this report. Morgan Stanley may have a position in the debt of the Company or instruments discussed in this report.
OTHER IMPORTANT DISCLOSURES The securities/instruments discussed in this report may not be suitable for all investors. This report has been prepared and issued by Morgan Stanley primarily for distribution to market professionals and institutional investor clients. Recipients who are not market professionals or institutional investor clients of Morgan Stanley should seek independent financial advice prior to making any investment decision based on this report or for any necessary explanation of its contents. This report does not provide individually tailored investment advice. It has been prepared without regard to the individual financial circumstances and objectives of persons who receive it. Morgan Stanley recommends that investors independently evaluate particular investments and strategies, and encourages investors to seek the advice of a financial advisor. The appropriateness of a particular investment or strategy will depend on an investor’s individual circumstances and objectives. You should consider this report as only a single factor in making an investment decision. Morgan Stanley fixed income research analysts, including those principally responsible for the preparation of this research report, receive compensation based upon various factors, including quality, accuracy and value of research, firm profitability or revenues (which include fixed income trading and capital markets profitability or revenues), client feedback and competitive factors. Analysts’ compensation is not linked to investment banking or capital markets transactions performed by Morgan Stanley or the profitability or revenues of particular trading desks. This report is not an offer to buy or sell any security/instrument or to participate in any trading strategy. In addition to any holdings disclosed in the section entitled “Important Disclosures on Subject Companies,” Morgan Stanley and/or its employees not involved in the preparation of this report may have investments in securities/instruments or derivatives of securities/instruments of companies mentioned in this report, and may trade them in ways different from those discussed in this report. Derivatives may be issued by Morgan Stanley or associated persons. Morgan Stanley makes every effort to use reliable, comprehensive information, but we make no representation that it is accurate or complete. We have no obligation to tell you when opinions or information in this report change. With the exception of information regarding Morgan Stanley, reports prepared by Morgan Stanley research personnel are based on public information. Facts and views presented in this report have not been reviewed by, and may not reflect information known to, professionals in other Morgan Stanley business areas, including investment banking personnel. The value of and income from investments may vary because of changes in interest rates, foreign exchange rates, default rates, prepayment rates, securities/instruments prices, market indexes, operational or financial conditions of companies or other factors. There may be time limitations on the exercise of options or other rights in securities/instruments transactions. Past performance is not necessarily a guide to future performance. Estimates of future performance are based on assumptions that may not be realized. This report may include research based on technical analysis. Technical analysis is generally based on the study of trading volumes and price movements in an attempt to identify and project price trends. Technical analysis does not consider the fundamentals of the underlying issuer or instrument and may offer an investment opinion that conflicts with other research generated by Morgan Stanley. Investors may consider technical research as one input in formulating an investment opinion. Additional inputs should include, but are not limited to, a review of the fundamentals of the underlying issuer/security/instrument. To our readers in Taiwan: Information on securities/instruments that trade in Taiwan is distributed by Morgan Stanley & Co. International Limited, Taipei Branch (the “Branch”). Such information is for your reference only. The reader should independently evaluate the investment
risks and is solely responsible for their investment decisions. This publication may not be distributed to the public media or quoted or used by the public media without the express written consent of Morgan Stanley. Information on securities/instruments that do not trade in Taiwan is for informational purposes only and is not to be construed as a recommendation or a solicitation to trade in such securities/instruments. The Branch may not execute transactions for clients in these securities/instruments. To our readers in Hong Kong: Information is distributed in Hong Kong by and on behalf of, and is attributable to, Morgan Stanley Dean Witter Asia Limited as part of its regulated activities in Hong Kong. If you have queries concerning this publication, please contact our Hong Kong sales representatives. Certain information in this report was sourced by employees of the Shanghai Representative Office of Morgan Stanley Dean Witter Asia Limited for the use of Morgan Stanley Dean Witter Asia Limited. THIS PUBLICATION IS DISSEMINATED IN JAPAN BY MORGAN STANLEY JAPAN SECURITIES CO., LTD.; IN HONG KONG BY MORGAN STANLEY DEAN WITTER ASIA LIMITED; IN SINGAPORE BY MORGAN STANLEY DEAN WITTER ASIA (SINGAPORE) PTE., REGULATED BY THE MONETARY AUTHORITY OF SINGAPORE, WHICH ACCEPTS RESPONSIBILITY FOR ITS CONTENTS; IN AUSTRALIA BY MORGAN STANLEY DEAN WITTER AUSTRALIA LIMITED A.B.N. 67 003 734 576, A LICENSED DEALER, WHICH ACCEPTS RESPONSIBILITY FOR ITS CONTENTS; IN CANADA BY MORGAN STANLEY CANADA LIMITED, WHICH HAS APPROVED OF, AND HAS AGREED TO TAKE RESPONSIBILITY FOR, THE CONTENTS OF THIS PUBLICATION IN CANADA; IN GERMANY BY MORGAN STANLEY BANK AG, FRANKFURT AM MAIN, REGULATED BY BUNDESANSTALT FUER FINANZDIENSTLEISTUNGSAUFSICHT (BAFIN); IN SPAIN BY MORGAN STANLEY, S.V., S.A., A MORGAN STANLEY GROUP COMPANY, WHICH IS SUPERVISED BY THE SPANISH SECURITIES MARKETS COMMISSION (CNMV) AND STATES THAT THIS DOCUMENT HAS BEEN WRITTEN AND DISTRIBUTED IN ACCORDANCE WITH THE RULES OF CONDUCT APPLICABLE TO FINANCIAL RESEARCH AS ESTABLISHED UNDER SPANISH REGULATIONS; IN THE UNITED KINGDOM BY MORGAN STANLEY & CO. INTERNATIONAL LIMITED, AUTHORISED AND REGULATED BY THE FINANCIAL SERVICES AUTHORITY; AND IN THE UNITED STATES BY MORGAN STANLEY & CO. INCORPORATED AND MORGAN STANLEY DW INC., WHICH ACCEPT RESPONSIBILITY FOR ITS CONTENTS. IN AUSTRALIA, THIS REPORT, AND ANY ACCESS TO IT, IS INTENDED ONLY FOR “WHOLESALE CLIENTS” WITHIN THE MEANING OF THE AUSTRALIAN CORPORATIONS ACT. NOT FOR DISTRIBUTION TO PRIVATE CUSTOMERS AS DEFINED BY THE UK FINANCIAL SERVICES AUTHORITY. The trademarks and service marks contained herein are the property of their respective owners. Third-party data providers make no warranties or representations of any kind relating to the accuracy, completeness, or timeliness of the data they provide and shall not have liability for any damages of any kind relating to such data. This report may not be sold or redistributed without the written consent of Morgan Stanley. Morgan Stanley research is disseminated and available primarily electronically, and, in some cases, in printed form. Additional information is available upon request. 4/19/06 lw
Structured Credit Insights
Contacts
Sivan Mahadevan (212) 761-1349
[email protected]
Peter Polanskyj (212) 761-1233
[email protected]
Vishwanath Tirupattur (212) 761-1043 [email protected] Pinar Onur (212) 761-1597
[email protected]
Andrew Sheets (212) 761-2341
[email protected]
THE AMERICAS 1585 Broadway New York, NY 10036-89293 Tel: (1) 212-761-4000 EUROPE 25 Cabot Square, Canary Wharf London E14 4QA, England Tel: (44 20) 7425-8000 JAPAN 20-3, Ebisu 4-chome, Shibuya-ku Tokyo 150-6008, Japan Tel: (813) 5424-5000 ASIA PACIFIC Three Exchange Square Hong Kong Tel: (852) 2848-5200
© 2006 Morgan Stanley
