Comments on:“Did Subjectivity Play a Role in CDO
Credit Ratings?”gPrepared for the Federal Reserve Bank of Atlanta Financial Markets Conference
May 11 2010May 11, 2010
by
Joseph R. MasonJoseph R. MasonProfessor and Hermann Moyse, Jr./Louisiana Bankers Association
Endowed Professor of Banking, Louisiana State University and
S i F ll Th Wh S h lSenior Fellow, The Wharton School
Contact [email protected], (202) 683-8909 office. Copyright Joseph R. Mason, 2010. All rights reserved.
Did Subjectivity Play a Role in CDO Credit Ratings?
but not the way the authors are addressing…but not the way the authors are addressing.
Purpose of SecuritizationPurpose of Securitization
P f S iti tiPurpose of Securitization
P f CDOPurpose of CDOs
P f M i ABS CDOPurpose of Mezzanine ABS CDOsABS #1 ABS CDOABS #1AAA/Aaa
AA/Aa2
A/A2
ABS CDOCollateral Pool
ABS #2AAA/Aaa
ABS CDO
Class A
BBB/Baa2Residual
BBB/Baa #2
AA/Aa2
A/A2BBB/Baa2Residual
ABS #3AAA/Aaa
BBB/Baa #1
Class B
Cl C
A/A2 #3AA/Aa2
A/A2BBB/Baa2Residual
ABS #4 BBB/Baa #4Class CClass DEquity
AAA/Aaa
AA/Aa2
A/A2BBB/Baa2
ABS #5AAA/Aaa
BBB/Baa #5
BBB/Baa2Residual AA/Aa2
A/A2BBB/Baa2Residual
Subprime RMBS Compositions in CDOs Subprime RMBS Compositions in CDOs Grew Quickly over Recent Years
Source: Fitch IBCA (2007)
Ratings Distribution of RMBS in CDO Ratings Distribution of RMBS in CDO Portfolios
Source: Fitch IBCA (2007)
P f CDOPurpose of CDOs
Pre-crisis CDOs Evaluated Pre crisis CDOs Evaluated Rationally
Loans Repurchased and Substituted in Private-label Mortgage Pools
600,000,000
700,000,000
Loans Repurchased and Substituted in Private label Mortgage Pools
400,000,000
500,000,000 Subjectivity?200,000,000
300,000,000
-
100,000,000
5 5 5 5 5 5 6 6 6 6 6 6 7 7 7 7 7 7 8 8 8 8 8 8 9 9
1/1/
2005
3/1/
2005
5/1/
2005
7/1/
2005
9/1/
2005
11/1
/200
5
1/1/
2006
3/1/
2006
5/1/
2006
7/1/
2006
9/1/
2006
11/1
/200
6
1/1/
2007
3/1/
2007
5/1/
2007
7/1/
2007
9/1/
2007
11/1
/200
7
1/1/
2008
3/1/
2008
5/1/
2008
7/1/
2008
9/1/
2008
11/1
/200
8
1/1/
2009
3/1/
2009
Source: BLIS, Blackbox Logic, June 2009 Remittance Report
CDO StructuresCDO Structures
CDO Hi hli htCDO Highlights• CDO is like a mutual fund that sells debt and • CDO is like a mutual fund that sells debt and
equity– CDOs have multi-layered capital structure with
several debt "tranches"• Early CDOs were backed by portfolios of junk
bondsbonds– Later CDOs emphasized corporate loans or ABS
• Portfolios are actively managed, subject to rulesy g , j• Diversification of the underlying portfolio is the
key to understanding the risk in a CDO
St t Additi l F tStructure – Additional Features• CDO lifecycleCDO lifecycle
– Ramp-up phase– Revolving phase– Amortization phase– Amortization phase
• Waterfall– Pre-2005: mostly sequential
P t 2005 tl t ( ti ith t l )– Post-2005: mostly pro rata (sometimes with toggle)• Collateral quality tests (eligibility)• Performance tests
– Overcollateralization (OC) – par haircuts– Interest coverage (IC)
• Events of Default
B i CDO St tBasic CDO Structure
Oth St tOther StructuresThis is the kind of structure everyone thinks of…
This has three s as ee AAA tranches, but the pricing suggests each has very different risk…
This has two or th AAA three AAA tranches, depending on who you listen to…
Source: Adelson, Mark. CDO/CDS Update 7/2/07, Nomura Fixed Income Research.
HEL ABS D l S C h FlHEL ABS Deal Structure – Cash Flow
• Senior-sub O/C (not prime MBS six pack)Senior sub, O/C (not prime MBS six pack)• Sequential / pro-rata / reverse sequential, with triggers• Importance of Eric Higgins, Joseph Mason, Adi
M d l “Th I f ti C t t f A t B k d Mordel, “The Information Content of Asset Backed Securities Downgrades and the Motivation behind Them” (2010)
Residual B Prin.
M Prin. A Prin.
0 12 24 36 48 60 72 84 96 108 120
Oth St tOther StructuresY Structures
H StructuresH Structures
Combination StructuresCombination Structures
St t Additi l F tStructure – Additional Features• CDO lifecycleCDO lifecycle
– Ramp-up phase– Revolving phase– Amortization phase– Amortization phase
• Waterfall– Pre-2005: mostly sequential
P t 2005 tl t ( ti ith t l )– Post-2005: mostly pro rata (sometimes with toggle)• Collateral quality tests (eligibility)• Performance tests
– Overcollateralization (OC) – par haircuts– Interest coverage (IC)
• Events of Default
E t f D f ltEvents of Default• OC test failures from collateral downgrades• OC test failures from collateral downgrades• Who gets to decide?• What are the choices?• References:
– Jordan P., et al., S&P Offers Guidance on Treatment of CDOs with O/C Events of Default; Two Ratings on W t h N S&P (10/31/07)
Subjectivity?Watch Neg, S&P (10/31/07)
– Kharnak, L., et al., Impact of Subprime Downgrades on OC-Linked Events of Default in CDOs, Moody's (11/1/07)(11/1/07)
– Lioce, S., Understanding the Consequences of ABS CDO Events of Default Triggered by Loss of Overcollateralization, Moody's (1/7/08)
Evolution of CDO ModelingEvolution of CDO Modeling
R ti A A hRating Agency Approaches• Monte Carlo simulation based approaches• Monte Carlo simulation-based approaches• Assumptions
– default frequencies– recovery rates– correlations
• Factor-based correlation modelsFactor based correlation models– Systems for assigning pair-wise correlations
• Software tools:– S&P: CDO Evaluator™ 3.3– Moody's: CDOROM™– Fitch: VECTOR 3.0
Source: Mark Adelson, “Collateralized Debt Obligations and Their Connection to Sub-prime Mortgages,” LexisNexis/Mealeys Conference Presentation, 6 March 2008
S&P M th d l E l ti (1)S&P Methodology Evolution (1)• Grossman R et al "High Yield Cash Flow • Grossman, R., et al., High Yield Cash Flow
Criteria," S&P CreditWeek, p. 19 (16 May 1988)– Initial move into rating CBOs of junk bonds– Rules developed from analysis of hypothetical
historical pool• Global CBO/CLO Criteria (1999)• Global CBO/CLO Criteria (1999)
– No explicit treatment of correlation; instead, focus on "diversification" and "risk concentration"
– Baseline industry concentration of 8%– Excess concentrations addressed by higher default
rate assumptionrate assumption
Source: Mark Adelson, “Collateralized Debt Obligations and Their Connection to Sub-prime Mortgages,” LexisNexis/Mealeys Conference Presentation, 6 March 2008
S&P M th d l E l ti (2)S&P Methodology Evolution (2)• Bergman S "CDO Evaluator Applies Correlation • Bergman, S., CDO Evaluator Applies Correlation
and Monte Carlo Simulation to the Art of Determining Credit Quality" (12 Nov 2001)– Introduction of Monte Carlo simulation approach Introduction of Monte Carlo simulation approach
through CDO Evaluator™– Corporate correlation: 30% intra-industry, 0% inter-
industry• 0% inter-industry assumption sharply criticized
(e.g., Cifuentes A., and N. Chen, "The Young and the Restless: Correlation Drama at the Big Three Rating Agencies," Wachovia Securities (22 Feb 2005))g ( ))
– ABS correlation: 30% intra-sector, 10% inter-sector– Produced lower stressed default rates for ABS than the
earlier Risk Tabulator model
Source: Mark Adelson, “Collateralized Debt Obligations and Their Connection to Sub-prime Mortgages,” LexisNexis/Mealeys Conference Presentation, 6 March 2008
S&P M th d l E l ti (3)S&P Methodology Evolution (3)"C it i F R ti S th ti CDO • "Criteria For Rating Synthetic CDO Transactions" (Sep 2003)– Heavy focus on documentation for synthetics– Heavy focus on documentation for synthetics– Specific treatment of synthetic CDOs of ABS (pp.
57-58)• credit events• settlement mechanisms (i.e., physical and cash)• recoveriesrecoveries
Source: Mark Adelson, “Collateralized Debt Obligations and Their Connection to Sub-prime Mortgages,” LexisNexis/Mealeys Conference Presentation, 6 March 2008
S&P M th d l E l ti (4)S&P Methodology Evolution (4)• Parisi F "Loss Correlations Among U S
Results: Loss Correlations by SF Asset Type Combinations
Parisi, F., Loss Correlations Among U.S. Consumer Assets" (Feb 2004)
(with 95% confidence intervals)
Asset Mfd. Hsg Bank Cards Auto Loans RMBS
0 55Mfd. Hsg. 0.55(0.518, 0.568)
Bank Cards 0.22(0.208, 0.233)
0.17(0.162, 0.179)
Auto Loans 0.37(0.356, 0.390)
0.21(0.201, 0.219)
0.48(0.464, 0.492)
RMBS 0.13 0.07 0.18 0.06RMBS (0.124, 0.133) (0.069, 0.076) (0.172, 0.185) (0.061, 0.064)
Source: Mark Adelson, “Collateralized Debt Obligations and Their Connection to Sub-prime Mortgages,” LexisNexis/Mealeys Conference Presentation, 6 March 2008
S&P M th d l E l ti (5)S&P Methodology Evolution (5)• Bradley E et al "CDO Spotlight: • Bradley, E., et al., "CDO Spotlight:
Synthetic CDO of ABS Documents Evolving Towards a Standard But Nuances Evolving Towards a Standard But Nuances Remain" (26 Apr 2005)– Published a few months before the release of
the first ISDA forms for CDS of ABS– Survey of documentation features
Identified varying practices in credit events – Identified varying practices in credit events and valuation concepts for settlements
– No mention of PAUG structures
Source: Mark Adelson, “Collateralized Debt Obligations and Their Connection to Sub-prime Mortgages,” LexisNexis/Mealeys Conference Presentation, 6 March 2008
S&P M th d l E l ti (6)S&P Methodology Evolution (6)• Gilkes K N Jobst and B Watson "CDO • Gilkes, K., N. Jobst, and B. Watson, CDO
Evaluator Version 3.0: Technical Document" (19 Dec 2005)
R l th ti f 0% i t i d t – Replaces the assumption of 0% inter-industry correlation for corporate bonds
– Produced watchlisting of 35 tranches from 18 synthetic CDO deals
Subjectivity?synthetic CDO deals
• 14 of the 18 deals carried ratings only from S&P• rating shopping issue
Reduced assumed intra industry corporate – Reduced assumed intra-industry corporate correlation to 15%
– Adopts different default rate assumptions for different types of instrumentsdifferent types of instruments
Source: Mark Adelson, “Collateralized Debt Obligations and Their Connection to Sub-prime Mortgages,” LexisNexis/Mealeys Conference Presentation, 6 March 2008
Global Default ProbabilitiesGlobal Default Probabilities
Ratings Default ProbabilitiesCantor and Packer (1996) showed that ratings for SF were more liberal than those for corporates.
AA+ CDO“…in December 2005, when S&P announced the release of version 3.0 of its CDO Evaluator software,
AA- ABSAA+ CDO
the agency published complete tables of default probabilities for ABS/MBS, corporate bonds, and CDOs. In doing so, it created g ,conflicting definitions for its rating symbols depending on the types of instruments to which they apply ”they apply.
Source: Standard & Poor's CDO Evaluator ver. 3.2 as reported in Adelson, Mark, “Bond Rating Confusion,” Nomura Fixed Income Research, June 29, 2006.
Ratings Default Probabilities“…on June 19, [2006] S&P released a new version of its CDO E l ™ Evaluator™ software, including new rating definitions for ABS As part of the ABS. As part of the software release, S&P changed the idealized default probabilities for default probabilities for ABS at different rating levels over different time horizons.”
Source: Standard & Poor's CDO Evaluator ver. 3.2 as reported in Adelson, Mark, “Bond Rating Confusion,” Nomura Fixed Income Research, June 29, 2006.
Ratings Bias v Ratings Quality
Charles W. Calomiris and Joseph R. Mason, “Conflicts of Interest, Low-Quality Ratings, and Meaningful Reform: Evidence
from Debt and Corporate Governance Ratings ” from Debt and Corporate Governance Ratings, (2010)
(policy draft at e21.com)
Eff f C fli f IEffects of Conflicts of Interest• Infl ted r tin s l r l t r• Inflated ratings loosen regulatory
constraints, which could arise either from attempts to maximize portfolio value or from aattempts to maximize portfolio value or from a desire to deceive clients about risk.
• Low-quality ratings arise primarily from theLow quality ratings arise primarily from the desire of investors to deceive clients and thereby invest too much in risky assets on behalf of ytheir clients in order to earn higher asset management fees
Source: Charles W. Calomiris and Joseph R. Mason, “Conflicts of Interest, Low-Quality Ratings, and Meaningful Reform: Evidence from Debt and Corporate Governance Ratings,” (2010)
Potential Sources of Conflict of Potential Sources of Conflict of Interest
Rating Agencies Institutional Investors
Ultimate Investors/Clients
•Preference for meaningful signal
•Preference for meaningful (?)
•Preference to produce meaningful signal
of investment quality.•Preference for
meaningful (?) signal of investment quality.
produce meaningful (?) signal of investment
risk-adjusted return
Regulatory arbitrage
Increased fees
quality.Issuer paysResponds to
institutional institutional investor demand
Source: Charles W. Calomiris and Joseph R. Mason, “Conflicts of Interest, Low-Quality Ratings, and Meaningful Reform: Evidence from Debt and Corporate Governance Ratings,” (2010)
Potential Sources of Conflict of Potential Sources of Conflict of Interest
We conclude that for both debt and corporate We conclude that, for both debt and corporate governance ratings, the low quality of ratings is largely
the result of the principal-agent conflict between institutional investors and ultimate investors.
Rating Agencies Institutional Investors
Ultimate Investors/ClientsSubjectivity?
•Potentially •Potentially •Potentially motivated by:
Plausible deniability
•Potentially motivated by
Shirking Institutional
Regulatory arbitrage
Increased fees
investor demand.
Source: Charles W. Calomiris and Joseph R. Mason, (2010)
ConclusionsConclusions
Conclusions and Conclusions and Recommendations
R• Recourse• Structure Details• Modeling Techniques• Default Probability BiasDefault Probability Bias
Subjectivity?Subjectivity?
Comments on:“Did Subjectivity Play a Role in CDO
Credit Ratings?”gPrepared for the Federal Reserve Bank of Atlanta Financial Markets Conference
May 11, 2010y
by
Joseph R. MasonJoseph R. MasonProfessor and Hermann Moyse, Jr./Louisiana Bankers Association
Endowed Professor of Banking, Louisiana State University and
S i F ll Th Wh S h lSenior Fellow, The Wharton School
Contact [email protected], (202) 683-8909 office. Copyright Joseph R. Mason, 2010. All rights reserved.
AppendixAppendix
A l C h CDO IAnnual Cash CDO Issuance500
400
450
500
250
300
350
150
200
0
50
100
1987 1989 1991 1993 1995 1997 1999 2001 2003 2005
Source: Lucas, Goodman, and Fabozzi (2006)
Moody's Methodology Evolution Moody s Methodology Evolution (1)
• Lucas D N Kirnon and L Moses "Rating Cash • Lucas, D., N. Kirnon, and L. Moses, Rating Cash Flow Transactions Backed by Corporate Debt" (Mar 1991)– Original methodologyOriginal methodology– Based on "WARF" and "Diversity Score"– Credit enhancement from tables– 32 industry classifications– 32-industry classifications
• Backman, A. and G. O'Connor, "Rating Cash Flow Transactions Backed by Corporate Debt, 1995 Update" (7 Apr 1995)Update (7 Apr 1995)– Essentially the same methodology– Addressed additional situations (estimated and implied
ratings)ratings)
Source: Mark Adelson, “Collateralized Debt Obligations and Their Connection to Sub-prime Mortgages,” LexisNexis/Mealeys Conference Presentation, 6 March 2008
Moody's Methodology Evolution Moody s Methodology Evolution (2)
Moody's Diversity Scoring
Firms in Same Industry Diversity Score
Moody's Rating Factors
Rating Factor Rating Factor
1 1.00
2 1.50
3 2.00
Aaa 1 Baa3 610
Aa1 10 Ba1 940
Aa2 20 Ba2 1,3504 2.33
5 2.67
6 3.00
7 3 25
Aa2 20 Ba2 1,350
Aa3 40 Ba3 1,780
A1 70 B1 2,2207 3.25
8 3.50
9 3.75
10 4 00
A2 120 B2 2,720
A3 180 B3 3,490
Baa1 260 Caa 6,50010 4.00
>10 Case-by-case
,
Baa2 360 Ca 10,000
Source: Backman, A. and G. O'Connor, "Rating Cash Flow Trans-actions Backed by Corporate Debt, 1995 Update" (7 Apr 1995)
Moody's Methodology Evolution (3)Moody's Methodology Evolution (3)• Cifuentes A. and G. O'Connor, "The Binomial Expansion
Method Applied to CBO/CLO Analysis" (13 Dec 1996)– Introduced "binomial expansion technique"– assumptions: default probability, diversity score, recovery rate– Formula for calculation of expected loss
D! p = g b bilit f d f ltjDjj pp
jDjDP −−−
= )1()!(!
! p = average probability of default
D = diversity score
j = no. of defaults in jth scenario
∑=
=D
jjjEP
1Loss Expected
j no. of defaults in j scenario
Pj= probability of scenario j
Ej= expected loss in scenario j=j 1
Source: Mark Adelson, “Collateralized Debt Obligations and Their Connection to Sub-prime Mortgages,” LexisNexis/Mealeys Conference Presentation, 6 March 2008
Moody's Methodology Evolution Moody s Methodology Evolution (4)
• Cifuentes A and C Wilcox "The Double • Cifuentes, A. and C. Wilcox, The Double Binomial Method and Its Application to a Special Case of CBO Structures" (20 Mar 1998)
Ad t ti f BET t i l – Adaptation of BET to special cases– DBM offers better view of certain pools such as 80%
emerging market debt combined with 20% U.S. high yield bank loanshigh yield bank loans
– Difference between BET and DBM is important only for low diversity pools (D<10) composed of two distinct asset groups (barbelled assets)distinct asset groups (barbelled assets)
– Example: two uncorrelated groups of assets with markedly different average properties
Source: Mark Adelson, “Collateralized Debt Obligations and Their Connection to Sub-prime Mortgages,” LexisNexis/Mealeys Conference Presentation, 6 March 2008
Moody's Methodology Evolution Moody s Methodology Evolution (5)
• Gluck J and H Remeza "Moody's Approach to Rating Gluck, J. and H. Remeza, Moody s Approach to Rating Multisector CDOs" (15 Sep 2000)– Introduced "alternative diversity score methodology" for
assets with correlated default risk
⎟⎠
⎞⎜⎝
⎛⎟⎠
⎞⎜⎝
⎛ ∑∑==
n
iii
n
iii n
FqFpD ifd11
– Correlation parameters disseminated confidentially
ρρρρ
=−+
=⎠⎝⎠⎝= ΣΣ==
ijjijjiiij
ii
nFFqpqpD if
)1(1and11
– Correlation parameters disseminated confidentially– Covered structured finance assets in addition to corporates– Addressed geographic, servicer, and vintage concentrations
Source: Mark Adelson, “Collateralized Debt Obligations and Their Connection to Sub-prime Mortgages,” LexisNexis/Mealeys Conference Presentation, 6 March 2008
Moody's Methodology Evolution Moody s Methodology Evolution (6)
• Tolk J "Understanding the Risks in Credit Default • Tolk, J., Understanding the Risks in Credit Default Swaps" (16 Mar 2001)– Initial focus on ISDA credit events for corporates– Comparison to Moody's default definition– Comparison to Moody s default definition
• issue of "soft" (non-default) credit events– Valuation and settlement– Brief mention of non-corporate reference creditsBrief mention of non-corporate reference credits
• Yoshizawa, Y., "Moody's Approach to Rating Synthetic CDOs" (29 July 2003)
Focus on multiple binomial method– Focus on multiple binomial method– Par per diversity test– Soft credit event update (stress default probabilities 5%
to 12 5%)to 12.5%)
Source: Mark Adelson, “Collateralized Debt Obligations and Their Connection to Sub-prime Mortgages,” LexisNexis/Mealeys Conference Presentation, 6 March 2008
Moody's Methodology Evolution Moody s Methodology Evolution (7)
• Witt G "Moody's Correlated Binomial Default ρWitt, G., Moody s Correlated Binomial Default Distribution" (10 Aug 2004)– Alternative to diversity score for capturing correlation– Assumes single default correlation between all asset pairs
ρg
– Also assumes default prob., recovery rate, & no. of assets– Closed form solution but computationally hard
• Fu, Y., et al., "Moody's Revisits Its Assumptions Regarding C D f l ( d A ) C l i f CDO (30 Corporate Default (and Asset) Correlations for CDOs (30 Nov 2004)– Inter-industry asset correlation 3%
Intra industry asset correlation 15%– Intra-industry asset correlation 15%• telecom and utilities 20%• chemicals, electronics, retail, textiles 10%
Source: Mark Adelson, “Collateralized Debt Obligations and Their Connection to Sub-prime Mortgages,” LexisNexis/Mealeys Conference Presentation, 6 March 2008
Moody's Methodology Evolution Moody s Methodology Evolution (8)
• Toutain O et al "Moody's Revisits Its Assumptions Toutain, O., et al., Moody s Revisits Its Assumptions Regarding Structured Finance Default (and Asset) Correlations for CDOs (27 July 2005)– "Tree" approach to sector risk: global, meta, broad, narrow– Add-ons for regional, vintage, and "key agent" effects– Complicated…
• Xie, M. and G. Witt, "Moody's Modeling Approach to Rating St t d Fi C h Fl CDO T ti " (26 S Structured Finance Cash Flow CDO Transactions" (26 Sep 2005)– Use CBM with single correlation parameter from CDOROM™– Assumed pair-wise correlations generate the single Assumed pair wise correlations generate the single
correlation parameter within CDOROM™• simulates losses• applies moment matching scheme to match skew of CBM loss
distribution to that of CDOROM™ simulated loss distributiondistribution to that of CDOROM simulated loss distribution
Source: Mark Adelson, “Collateralized Debt Obligations and Their Connection to Sub-prime Mortgages,” LexisNexis/Mealeys Conference Presentation, 6 March 2008
Moody's Methodology Evolution Moody s Methodology Evolution (9)
• Kim T "Moodys Initial Views on the Dealer Form of Kim, T., Moodys Initial Views on the Dealer Form of Confirmation for Pay-As-You-Go Derivative Transactions" (21 Jun 2006)
• Bharwani, P., "Moody's Approach to Rating Collateralized Debt Obligations with Pay-As-You-Go Credit Default Swaps" (13 Nov 2006)– Objects to appraisal reduction as writedown
Strongly objects to implied writedowns but allows as floating – Strongly objects to implied writedowns but allows as floating payment event if certain conditions satisfied
– Rating downgrade to Caa OK w/ physical settlement only• Ca cash settle OK after six monts• C cash settle OK immediately
– Counterparty risks– Amendments
Source: Mark Adelson, “Collateralized Debt Obligations and Their Connection to Sub-prime Mortgages,” LexisNexis/Mealeys Conference Presentation, 6 March 2008