ALLL Model Validation

John Hurlock

ALLL model validation has gained in emphasis and importance due to advances associated with new technology used to determine the ALLL, as a result of the banking crisis and in preparation of moving to an expected loss accounting model.

Can you have a valid model and still fail an audit? Yes.

  • A wrong decision could be made based on poor data.
  • The model or its results could be misapplied.
  • A lack of trust in the output leads to a delayed decision.
  • Missing or inaccurate information could lead to an invalid estimate.

The model validation process includes review of:

  • Purpose: the reason the model is employed including the intended use of the information extracted from the model
  • Inputs: automated and manual inputs as well as documented and undocumented
    assumptions used to create the model, including compliance requirements
  • Transformation: formulas used to transform data into information used to manage the activity
  • Outputs: the presentation of the information and the ability to interpret the results in a consistent fashion

Correlation is key to a valid model: how well the data correlates to the selected targets – national, local, peers – and the period of time being measured. A leading correlation is ALLL to total assets, funding compared to total charge-offs. Some models don’t look back far enough to demonstrate that information is handled accurately.

Common problems discovered during model validations:

  • Growing banks challenged by Excel models
  • Lack of project plans for converting to software
  • Too much reliance on software vendors for knowledge
  • Lack of a transition plan to a new model
  • Accounting for acquired loans
  • Lack of documented procedures

Keys to model risk management:

  • Establish an inventory of all models and rank by impact on the organization.
  • Formally explain why the model was created or purchased, and what the intended use of the model is, including if results are used as inputs into another model.
  • Determine how the model is used in decision-making and management.
  • Document to determine the likelihood of a negative event.
  • Ensure IT standards are maintained.

Looking ahead:

  • The banking crisis will continue to define approaches to the ALLL into the future.
  • The roles of capital and provision will become better defined.