Best practices in validating CECL models

Best practices in validating CECL models

By Jacob Kosoff, Head of Model Risk Management & Validation, Regions Bank.

Jacob, can you please tell the Risk Insights readers a little bit about yourself, your experiences and what your current professional focus is?

My professional focus is on using better modeling and advanced analytics to streamline processes and increase revenue opportunities.  I am the Head of Model Risk Management and Validation at Regions Bank and I have been in this role for about 4 years.  I have 15 years of experience in market risk and credit risk modeling and analytics, having worked at PNC, Freddie Mac, and Genesis Analytics prior to my current role at Regions bank.   My wife, three kids and I love living in Birmingham, have really settled into the city well and have become active in the civic life of Birmingham.


At the CECL Congress 2018, you will be speaking on your insight regarding “Best practices in validating CECL models’. Why is this a key talking point in the industry right now?

The key points in validating CECL models right now are model methodology, portfolio segmentation and data.  On the first point on methodology, areas of focus will include portfolio level modeling vs loan level modeling, as well as integration with upstream and downstream models.

What, in your opinion, are the limitations going forward?

A major limitation in modeling is the assumption of steady state portfolios.  As portfolios shrink or grow, the models must be able to adjust to a longer or shorter average time on book.

  How can financial institutions best handle model risk management policies and governance?

Financial institutions can best handle model risk management policies and governance by applying the principles of SR 11-7 to the entire CECL process.  One key first step in this process is to perform a skills assessment of the validation associates involved in the validation of the CECL model and ensure your team has the sufficient knowledge, skills and experience needed.  This will range from knowledge of accounting rules to experience with internal data sources to having sufficient mathematical and statistical skills.


Finally, what challenges do you foresee with CECL implementation over the coming years?

Institutions can best plan for the challenges and deadlines by kicking off validation of CECL models early and assigning multiple qualified validation associates on the project to ensure there is limited key-person dependency.  Key topics will have to be considered internally including how to treat TDRs, developing multiple scenarios for reasonable and supportable forecasts, and how best multiple stakeholders can coordinate to ensure a well-controlled implementation.