Corporate IRB Implementation: Overview from the U.S. Supervisory Perspective Advanced IRB Forum June 19, 2003

1
Corporate IRB Implementation Overview from
the U.S. Supervisory PerspectiveAdvanced IRB
Forum June 19, 2003
2
Purpose of Supervisory Guidance

• Indicate to industry and supervisors how IRB will
  be implemented in the United States.
• Allow banks to continue developing credit-risk
  management systems consistent with the IRB
  framework.
• Focus the dialogue between industry and
  supervisors on key issues.
• Ensure consistency across agencies in the
  implementation of supervisory initiatives related
  to IRB.
• Provide a foundation for staff development,
  training, and continuing refinement of
  supervisory policy.

3
Structure of the Guidance

• Guidance is organized into four chapters,
  reflecting the key elements of any IRB system
• Rating System - a system that assigns ratings
  and a process to validate the accuracy of those
  ratings
• Quantification - a quantification process that
  translates risk ratings into IRB parameters (PD,
  LGD, EAD, M)
• Data - a data maintenance system that supports
  the operation and refinement of the risk rating
  system
• Controls - oversight and control mechanisms that
  are designed to ensure the system is producing
  accurate and consistent ratings

4
Overarching Themes

• IRB Standards will help the industry understand
  supervisory expectations.
• However, application of these standards will
  require exercise of judgment on the part of
  supervisors.
• Training and consistency will be essential.
• Guidance will evolve over time, as we and the
  industry learn more about ratings based
  approaches to capital.
• Implementation will require extensive dialogue
  between banks and supervisors to understand,
  clarify and address key issues.
• A conservative bias will be necessary until
  adequate experience accumulates with the new
  approach.
• Ultimately, better alignment of risk and capital
  may permit reductions in minimum required capital
  at IRB banks.

5
Chapter 1 Rating System Design

• The first component of an IRB system involves the
  assignment and validation of ratings.
• Two-dimensional risk-rating system Meaningful
  and consistent differentiations among exposures
  along two dimensions, reflecting obligor default
  risk and loss severity.
• Calibration Obligor and loss severity ratings
  must be calibrated to probability of default (PD)
  and loss given default (LGD).
• Accuracy Actual long-run default frequencies
  for obligor grades close to PD estimates actual
  loss rates on defaults close to LGD estimates.
• Validation process Ongoing validation processes
  for rating systems must include (1) the review of
  developmental evidence, (2) ongoing monitoring,
  and (3) comparison of predictions to actual
  outcomes through back-testing.

6
Industry Challenges

• Systems Changes Many banks have recently
  revamped their rating systems to be
  two-dimensional others are preparing for this
  fundamental change. Most have little experience
  with this approach.
• Experts Versus Models Commonly used
  expert-judgment based systems may be used, but
  may face a challenging hurdle in meeting
  supervisory standards.
• Rating Philosophy Banks must more fully
  articulate their rating approach (not just
  point-in-time or through-the-cycle) and
  reflect that choice in other aspects of the
  rating system.
• Accuracy and Validation Banks must work to
  develop appropriate tests of ratings accuracy
  the exact nature will depend on details of each
  banks rating philosophy.

7
Chapter 2 Ratings Quantification

• IRB capital calculations require banks to supply
  four inputs or parameters.
• Probability of Default PD
• Loss Given Default LGD
• Exposure At Default EAD
• Maturity M
• Challenge for supervisors In practice, banks use
  many different approaches, estimation methods,
  and data sources.
• Our supervisory approach should be robust to
  these differences, and should not unduly
  constrain the industrys choice of methods and
  data when so little is known about many aspects
  of quantification.
• Solution guidance provides principles of sound
  practice, and a framework for supervisory
  dialogue with banks.

8
Quantification Framework

• The guidance presents a four-element framework
  that can be applied to virtually any approach to
  ratings quantification
• Data identify or construct a reference data
  source
• Estimation apply statistical techniques to the
  reference data to derive parameter estimates
• Mapping create a link between the reference data
  and a banks actual portfolio data
• Application apply parameter estimates to each
  exposure in the portfolio

9
Key Quantification Standards

• Specified Process Must have a fully specified
  process covering all four elements (data,
  estimation, mapping, and application).
• Complete Documentation Process must be fully
  documented, and must be updated periodically.
• Independent Review All aspects of the
  quantification process, including design and
  implementation, must be subject to an appropriate
  degree of independent review and validation.
• Constraints on Judgment Judgmental adjustments
  may be appropriate, but must not be biased in the
  direction of lower capital.
• Conservatism Parameter estimates must
  incorporate a degree of conservatism appropriate
  to the overall robustness of the process.
• Comprehensive Validation A validation process
  must cover all aspects of IRB quantification.

10
Industry Quantification Challenges

• Formalized process IRB banks will need to
  formalize their process by documenting their
  quantification process more thoroughly than
  currently done in practice.
• Reference data requirements Requirements for
  number of years and range of conditions covered,
  using all available data, and establishing
  comparability.
• Independent assessment Who will independently
  assess the work of the quants who produce the
  estimates?

11
Chapter 3 Data Maintenance

• Ability to implement IRB hinges on having
  adequate database
• Systems to fully implement Basel do not yet exist
  at most banks
• Lack of data has hindered full implementation of
  idealized risk management frameworks for many
  years
• Uncertainty
• IRB Framework has launched banks and supervisors
  into uncharted territory
• Banks and supervisors need more data to gain
  comfort than if these practices had been in place
  for a decade.
• Supervisors and risk managers should use data to
  gain comfort that ratings assignments and
  parameter estimates are reliable

12
Chapter 3 Data Maintenance -- Key Standards

• Collect Data Over Life of Loan
• cradle to grave collection of data
• obligors and facilities
• beyond the grave recoveries and costs.
• Collect Rating Assignment Data
• significant quantitative and qualitative factors
• obligors and facilities
• Support of IRB System -- data collected by
  institutions must be of sufficient depth, scope,
  and reliability to
• Validate IRB system processes,
• Validate parameters,
• Refine the IRB system,
• Develop internal parameter estimates,
• Apply improvements historically,
• Calculate capital ratios,
• Produce internal and public reports, and
• Support risk management.

13
Industry Data Challenges

• Sophisticated data maintenance will cost banks
  tens of millions of dollars
• New systems will require far reaching cultural
  and process changes
• Data systems must serve the needs of the line,
  risk managers, and control functions and not the
  other way around
• The selected IRB framework should dictate the
  content and structure of the system

14
Chapter 4 Control and Oversight

• Linking internal ratings to regulatory capital
  places new pressures on individuals and systems
• Banks are expected to have a strong system of
  controls for
• keeping incentive conflicts in check
• maintaining rating system integrity
• Scope of control and oversight mechanisms
• ratings assignment
• design of rating system
• quantification of ratings
• data management and integrity
• validation of rating system
• Fundamentals
• separation of duties
• balancing incentives
• layers of review

15
Key Supervisory Standards

• Interdependent System of Controls -- IRB
  institutions must implement a system of
  interdependent controls that include the
  following elements
• independence
• transparency
• accountability
• use of ratings
• rating system review
• internal audit
• board and senior management oversight
• Checks and Balances -- Institutions must combine
  the various control mechanisms in a way that
  provides checks and balances for ensuring IRB
  system integrity.

16
Control Challenges

• Banks will need to change the scope and focus of
  control functions
• creation of new rating systems review function
• work of high level risk managers and quants has
  not received intense scrutiny by qualified
  experts in a control function
• audit has not focused on technical quant issues
  and ratings have focused on problems, not pass
  grades
• Supervisors will need to conduct rigorous on-site
  reviews that evaluate a complex set of
  trade-offs, especially around ratings integrity.
• Supervisors will need to apply their judgement in
  evaluating how well checks and balances fit
  together.