Recalibration of CECL models after initial validation reports and parallel runs to align results

Recalibration of CECL models after initial validation reports and parallel runs to align results

By Elizae Dalvi, VP, Model Risk Management, Bank United

CECL 2019 is taking place in New York City on 27-28 March, 2019 – find out more here

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

What, for you, are the benefits of attending a conference like CECL? I am Vice President of Model Risk Management at BankUnited, Miami, FL and manage all aspects of model risk at the bank. We are currently conducting validations of enterprise-wide models in compliance with SR 7-11, which include a collection of PD/LGD/EAD models for CECL use. My background is in quantitative modeling, data analytics, statistics, and computer programming.
Attending a conference like CECL provides insights into how the industry is converging around key assumptions and approaches, given that the CECL guidelines are not prescriptive. It also provides an opportunity to learn from best practices.

How can qualitative overlays best be used to manage volatility?

Qualitative factors (Q factors) can be used to adequately fit historical information to current conditions which are not reflected in the model or economic scenarios used, such as the impact of changes in underwriting standards or local business conditions.

How can issues arising from parallel runs be fixed and how can firms detect areas that validation doesn’t pick up on?

It is expected that parallel testing will yield different results and this does not necessarily indicate a problem. However, it is important for model risk to follow the process all the way from model implementation to the parallel run to ensure that decisions regarding credit risk as well as accounting have been implemented correctly. Variance analysis can also be used to analyze and understand material differences in results.

In your opinion, what are the priorities to look out for when tweaking and recalibrating forward looking models?

It is important to use an out-of-sample testing period when gauging model performance and try not not to overfit models.

What challenges could firms expect to face when aligning model outputs with reasonable expectations?

The first challenge is determining the appropriate length of the reasonable and supportable period, which should consider the type of product and Its run-off patterns. The second is ensuring that the view on the economic environment incorporates both positive and negative economic developments in order to be “true and fair”.  And lastly, ensuring that the model performs well in benign economic environments.

How do you see the impact of CECL evolving over the next 6-12 months?

From a model risk perspective, in my opinion the focus will be around establishing sound governance, controls, and checking the reasonableness of results with parallel runs. It will also be impactful for model risk to ensure that the accounting decisions have been implemented properly, as some might remain outstanding.

You may also be interested in…

Sign up for your free Risk Insights account: