Brian D Gordon Director Brian gordonfitchratings com Overview

Brian D. Gordon, Director Brian. gordon@fitchratings. com

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

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

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

U. S. High Yield Default Index 1980 - 2002 120 Default Rate 80 60 40 20 0 The 2001 default rate excluding fallen angels was 9. 7% The 2002 default rate excluding fallen angels was 12. 4% Default Rates Default Volume (000's) 100 Default Volume

It’s 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

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

Drivers of Asset Default Rates § Rating of the underlying assets § Expected Life of the Assets § Correlation among the assets

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)

Cumulative Gross Default Rates

Marginal Gross Default Rates

Why Does Correlation Matter? 2001 2002 Total = $78. 2 bn Paper & Forest Products Industrial/ 2% Manufacturing 2% Total = $109. 8 bn Leisure & Food, Beverage & Tobacco Entertainment 2% 2% Transportation 2% Chemicals 3% Metals & Mining 3% Automotive 5% Telecommunication 36% Other 11% Banking & Finance 13% Utilities 19%

What is Correlation? § The degree to which two series of variables move in unison (1)

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)

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

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

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 Recovery § B BB BBB A AA AAA 65% 63% 60% 55% 50% 45% Recoveries will also be time lagged for cash deals, but not for synthetics because of immediate valuation procedures

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

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

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

The Impact of Correlation on Default Distribution

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”

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

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

www. fitchratings. com