An Overview of Fitchs new CDO Criteria Development

1
An Overview of Fitchs newCDO Criteria
Development

• Brian D. Gordon, Director
• Brian.gordon_at_fitchratings.com

2
Overview of Presentation

• Fitch is working on developing a completely new
  methodology for evaluating and rating CDOs
• The criteria is still a work in progress and
  subject to substantial change prior to release
• This presentation will give a sneak peak of the
  underlying approach, logic and application

3
The World in 1997

• 1997 was the first year that saw a substantial
  issuance of CDOs
• CDO rating methodologies created at about the
  same time by all three agencies
• The core of all three methodologies is basically
  unchanged from that time

4
But the world has changed substantially since
then

• Asian crisis in 1997
• Russian default plus Long Term Cap Mgmt 1998
• Bubble Economy 1999-2000
• Record High Yield Default Rates 2001-2002
• CDO new issuance volume reaches 85 billion in
  2002
• Market volatility increases dramatically

5
U.S. High Yield Default Index1980 - 2002
120
Default Volume
Default Rate
100
80
60
Default Rates
Default Volume (000's)
40
20
0
The 2001 default rate excluding fallen angels was
9.7 The 2002 default rate excluding fallen
angels was 12.4
6
Its Time for a Fresh Look at CDOs

• CDOs are among the most innovative and complex
  financial structures in existence
• CDO types include cash, synthetic, market value,
  high yield, high grade, trust preferred, etc.
• They are self-contained portfolios of credit risk
• The same construct that is applied to CDOs can be
  applied to any portfolio of credit risk,
  including ABCP, SIVs, bank and insurance
  portfolios

7
The Major Drivers of Risk in CDOs

• Default rates of underlying assets
• Recovery rates of underlying assets
• Structural considerations
• Interest rate risk, FX risk
• Management Risk, Moral Hazard
• Execution and ramp-up risk

8
Drivers of Asset Default Rates

• Rating of the underlying assets
• Expected Life of the Assets
• Correlation among the assets

9
Measuring Asset Default Risk

• Fitch will introduce an entirely new Default
  Matrix
• Based on empirical default rate evidence from all
  three agencies
• 30 year cohort analysis
• Establishes base case default expectations by
  rating (AAA to B) and life (1 to 10 years)

10
Cumulative Gross Default Rates
11
Marginal Gross Default Rates
12
Why Does Correlation Matter?
2001
2002
Total 78.2bn
Total 109.8bn
Food, Beverage
Leisure
Paper Forest
Tobacco
Entertainment
Products
Industrial/
2
2
2
Transportation
Manufacturing
2
2
Chemicals
3
Metals
Mining
3
Automotive
Telecommunication
5
36
Other
11
Banking
Finance
Utilities
13
19
13
What is Correlation?

• The degree to which two series of variables move
  in unison

14
What is Default Correlation?

• The degree to which the default probabilities of
  two firms move in unison
• A structural model model of default, based on
  the Black-Scholes option pricing model

(2)
(3)
(4)
15
The Correlation Matrix

• 25 Fitch defined industries
• All companies within an industry similarly
  correlated
• Inter-industry correlation is pair-wise
  correlation among industries (e.g. Chemicals to
  Auto)
• Intra-industry correlation is the correlation
  within an industry (e.g. Chemicals to Chemicals)
• Correlation for each company expressed as Sector
  Average Security which is a multiple regression
  across all other industries plus epsilon, a
  random variable representing unsystematic risk

16
Recovery Rates

• Recovery rates are a function of four variables
• 1) Systematic risk, implying that recoveries
  are inversely correlated to default rates
• 2) Idiosyncratic risk, meaning the unique
  properties of that company
• 3) The position of the debt in the capital
  structure of the company
• 4) The industry of the company

17
Fitch CDO Recovery Rate Matrix

• Fitch introduces the concept of tiered recovery
  rates, where the recovery rate varies with the
  stress scenario
• For example, US Senior secured bank loans
• B BB BBB A AA AAA
• Recovery 65 63 60 55 50 45
• Recoveries will also be time lagged for cash
  deals, but not for synthetics because of
  immediate valuation procedures

18
CDO Modeling

• The Monte Carlo Model
• Generates a vector of defaults and recoveries
  that are required for each rating level
• The Cash Flow Model
• Generates payment streams to rated liabilities
  using the payment waterfall and liability
  structure

19
Monte Carlo Simulation

• Uses a very large number of trial values for one
  or more random variables to produce a probability
  density function
• The brute force method to solving differential
  equations
• Used in the default generation model to produce
  inputs into the Cash Flow Model
• May be applied to the Cash Flow Model in the
  future as well

20
Monte Carlo Simulation

• Rating level default probability Di
• Random number generator Vi
• pi Pr(Vi Di)
• The degree in which two random variables are
  truly random, or conversely, move in unison, is
  dictated by the correlation assumption

21
The Impact of Correlation on Default Distribution
22
Default Distribution
N 100 P 10 (i I, , N)
Number of Defaults
96 C.I. C. G. D. R. 23 BBB
99.5 C.I. C. G. D. R. 40 AAA
23
Summary of the New Approach

• Draws upon empirical evidence for underlying
  assumptions about defaults, recoveries and
  correlation
• Employs a rigorous mathematical approach
• Uses state of the art modeling techniques,
  including Monte Carlo simulations
• Widely applicable to all types of CDOs plus other
  credit dependent portfolios

24
Roll Out and Impact

• Expected release late Spring 2003
• Immediate implementation after release
• Likely to be more conservative than existing
  criteria
• Fitch will release an article on the application
  of the new criteria to new and existing deals
• Same methodology will be applied in Europe

25
www.fitchratings.com