Case Study

A large North American credit union improves capital reserving process with ALLL model validation

  • Facebook
  • Twitter
  • Linkedin
  • Email

A large North American credit union

Financial services

Business need addressed:

  • The client had developed and proactively adopted a new capital plan adhering with recently published National Credit Union Administration (NCUA) guidelines
  • Genpact independently reviewed and assessed the model risk inherent for the client’s ALLL - Allowance for loan and lease losses model. The team highlighted key model shortcomings and suggested improvement areas

Genpact solution:

  • Checked the conceptual soundness and verified statistical accuracy and stability of the ALLL model and suggested potential corrective action plans that the client may need to undertake to address the inherent model risk
  • Defined relative severity specific to corrective action plans to assist the client in prioritizing the remediation activities

Business impact:

  • Full-scale model validation as per OCC SR 11-7 requirements, highlighting key model shortcomings
  • Model component level validation to gauge accuracy and appropriateness of projected and allotted reserves
  • Evaluated the extent of inherent model risk and its implication on ongoing usage of model outputs
  • Suggested improvement areas that require immediate to short term attention to ensure uninterrupted usage of the model output
  • Long term improvement actions to put in place a best-in-class as well as accurate reserving process

Model overview

ALLL model is an internally built spreadsheet that computes reserve (both gross and net) requirements for 12 months forward-looking period, both for homogenous and specific reserve pools, in line with ASC450 and ASC310 requirements respectively. Primary drivers for the model were found to be ex post facto loss rates, recoveries, loan balances, and macroeconomic indicators. The model is used for forecasting future loss rates and ALLL reserve levels by portfolio types as well as aggregated pool levels.

The model considers segregation of homogenous and specific pools each into a total of eight product lines as mentioned below:

  • Personal
  • Line of credit (LoC)
  • Non real estate secured
  • Business
  • Home equity
  • Credit cards
  • Mortgage
  • Student

Take a copy for yourself

Download PDF

The model embeds macroeconomic layering in the overall reserving process through a heuristic framework and uses national, state, and local level macroeconomic indicators.

Genpact independent review approach

As part of the review and validation process, model validation team (MVT) at Genpact evaluated the conceptual soundness of the model and performed various quantitative tests to gauge the model performance, stability, conservatism and ongoing applicability of the output.

MVT reviewed following areas to assess the methodological soundness of the ALLL model:

  • Quality of model design and construction: Reviewed various elements of model design and construction
  • Applicability as per intended use tests: Evaluated the fitment of the model for the intended use test purposes
  • Oversight and governance: Reviewed the availability of governance and oversight from board of directors and senior management as well as existence of policies and control mechanisms
  • Completeness of documentation: Reviewed and validated the availability, appropriateness, and completeness of documentation related to model development, initial validation, ongoing use etc.

For back testing purpose, MVT performed various quantitative tests to check:

  • Appropriateness of applying ex post facto loss rates as proxy for future loss rates
  • Comparative fitment of application of 3-months, 6-months or 1-year rolling average ex post facto loss rates at ASC450 and ASC310 pool levels
  • Conservatism in application of reserves
  • Accuracy of reserving process by comparing between predicted and actual losses over a period of time and across portfolio types
  • Appropriateness of considered macroeconomic factors in the reserving process

Basis the review findings and understanding of underlying complexity and materiality of the reserving process, the MVT rated the model to be at "High" risk.

Key observations and recommendations

As part of model validation process, MVT captured numerous findings and provided recommendations for correction and enhancement. Figure 1, outlines some of the key observations and their corresponding recommendations.


The importance of a full-scale independent review can be multifold. While it can provide a close view of the model risk embedded in the model output, it invariably allows the firm a chance to enhance the model to ensure a best-in-class as well as regulatory compliant process. For models such as reserving or capital computation models, rigor in model review and validation is essential to strike a balance between over-reserving and under-reserving. As part of this assignment Genpact noticed client's ALLL reserve model to satisfy the base requirements of the reserve computation process. However, it was found to be lacking a futuristic predictive process, which is necessary for the allocated reserves to be attuned to portfolio dynamics and market movements. The review also indicated that although the design and assumptions of the model are conditionally appropriate for its intended purpose, there exist several thoughts to consider redevelopment of ALLL model by adopting advanced predictive methodologies such as ARIMA, ARIMAX, Dual-time Dynamics etc.

For more information, contact

Visit our Commercial Banking page