Peter O. Davis

1
Current State of Credit Risk Measurement
Symposium on Enterprise Wide Risk
Management Chicago, April 26, 2004

• Peter O. Davis
• Partner, Ernst Young LLP
• Director of Credit Risk Services
• peter.davis_at_ey.com

2
Agenda

• Continued Movement Toward Credit Quantification
• Regulatory Push
• Avoiding Unintended Consequences
• Incidence vs. Dollar Based Default Rates

3
Continued Movement Towards Credit Quantification

• Extending credit inherently a judgment-based
  decision
• Continued movement toward the reliance on credit
  models to support credit extension and portfolio
  management
• Driven in part by continued advances in credit
  risk modeling
• More mature models
• Greater computing power
• Development of credit loss databases
• More credit products providing market information
• Driven in part by demand for greater transparency
• Large defaults by fallen angels triggered
  increased focus by investors
• Demand for greater information by senior
  management, Board, shareholders, rating agencies,
  regulators
• Demand for consistent measurement across products

4
Regulatory Push

• For commercial banks, and (more recently)
  investment banks, regulators have created
  incentives for institutions to enhance their
  internal credit models
• For those meeting advanced standards, by year-end
  2006, under Basel II regulators will rely upon
  institutions internal credit models for setting
  regulatory capital
• Probability of default models
• Loss given default models
• Exposure at default models
• Will result in
• Standardization of credit risk measurement
  terminology and model classification
• Heavy focus on model accuracy
• Development of extensive credit performance
  databases, leading to ongoing innovations in
  model development
• Greater transparency in credit risk-taking across
  institutions
• Greater liquidity and continued innovation in
  credit products

5
Avoiding Unintended Consequences

• As credit models are used more broadly across
  institutions and more deeply within institutions,
  continued need to challenge whether
• models accuracy capture risks
• model limitations are understood
• application of individuals models and the
  integration of multiple models produce results
  that are consistent with the intended measurement
  purpose
• Example of the application of default models

6
Incidence-based vs. Dollar-based Defaults
Illustration

• Obligor default models measure the probability
  that an individual borrower will default over a
  given time horizon an incidence measure of
  default risk
• When measuring expected loss (EL), it is common
  to use the product of the probability of default
  (PD), loss given default (LGD) and exposure at
  default (EAD)
• This approach implicitly assumes that
  incidence-based and dollar-based PDs are the same

7
Illustration Contd

• Illustration of the impact of dollar-based vs.
  incidence-based default rates

• Assumptions
• Three banks with loan portfolio of 100 million
• Same 10 borrowers and 3 defaults
• Obligors have same risk rating and incidence
  based default rates
• LGD 100 for defaulted loans
• 100 closed-end loans

8
Illustration Contd

• Bank A Loan Size Does Not Differ
• All the loans are the same size
• The dollar loss implied by the incidence-based
  default rate is the same as historical loss
• Bank B Large Positions in Loans to Defaulting
  Obligors
• The dollar loss implied by the incidence-based
  default rate is based on the number of defaults
  and average value of the loans
• The size discrepancy of the loans are so large
  the average value hides more than it reveals
• Bank C Small Positions in Loans to Defaulting
  Obligors
• The dollar loss is far lower than the
  incidence-based default rates imply

9
Loss Severity Adjustment

• Loss Severity Adjustment is defined as the ratio
  of the average value of defaults to the average
  current balance of the portfolio
• For Bank B and C the use of incidence-based PD
  may not reflect the trends of the portfolio
• For such instances a Loss Severity Adjustment may
  be applied to the losses implied by the
  incidence-based default rates
• After the adjustment, the dollar losses reflect
  historical figures and the trend in the portfolio
  towards higher or lower dollar-based default
  probabilities
• Loss Severity Adjustment restores information
  lost in the averages by reconciling the
  incidence-based default rates with the banks loss
  in dollars