CECL model implementation and execution: 1st line model validation / risk management and running alongside stress tests in a cohesive framework

By Michael Jacobs, Lead Quantitative Analytics and Modeling Expert, PNC

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

I am a lead model development and analytics expert across a range of risk and product types, having a focus on wholesale credit risk methodology, regulatory solutions and model validation. At PNC I lead the 1st line model risk management and validation function for the C&I asset class. I am also an internal advisor on modelling methodologies and a liaison to industry groups, academia and the regulators. I have 25 years of experience in financial risk modeling and analytics, having worked 5 years at Accenture and Big 4 consulting as a Director in the risk modeling and analytics practice, with a focus on regulatory solutions; 7 years as a Senior Economist and Lead Modeling Expert at the OCC, focusing on ERM and Model Risk; and 8 years in banking as a Senior Vice-President at JPMC and SMBC, developing wholesale credit risk and economic capital models. Skills include model development & validation for CCAR, PPNR, CECL, credit / market / operational risk; Basel and ICAAP; model risk management; financial regulation; advanced statistical and optimization methodologies. I hold a doctorate in Mathematical Finance from the City University of New York – Zicklin School of Business and is a Chartered Financial Analyst. Mike has an extensive publication track record in several prestigious academic and practitioner journals, and has been a speaker at many high profile venues.

What, for you, are the benefits of attending a conference like CECL USA and what have attendees learnt from your session?

Attendees gained 1st hand insights in CECL implementation and execution from the perspective of a large bank and an experienced industry professional. This included key challenges, solutions and lessons learned in this space. Furthermore, having worked for several banks previous as a consultant, as well as a regulator involved with several banks prior to that, I gave an industry perspective regarding issues that are related to the current CECL initiative (i.e., stress testing, model risk management, supervisory expectations regarding models that serve regulatory purposes).

What, for you are some of the key considerations when transitioning to CECL and running alongside stress tests?

One key consideration of utmost importance in ensuring that models are implemented in line with the intent of model development. Another key consideration is having a a structured process for verifying and testing any changes to models or implementation code as we go through parallel runs. Furthermore, it is critical that stakeholders have an understanding of results from a reasonableness perspective, and this should be informed by the expertise of model developers, in order that we can distinguish model from non-model impacts upon results.

Please describe some of the challenges involved with interaction between IT and model development?

A key challenge is in the process of translating modelling domain knowledge into a form that is understandable to technologists. An example is in the differences in the forms of implementation testing that model development vs. IT consider, things like regression and systems integration in the former, and sensitivity and attribution analysis for the latter. Another important challenge is around reconciling timelines, as the requirements of production may differ from that of implementation of development model prototypes.

In your opinion, what key considerations need to be made in macroeconomic variables & scenario generation?

The most important consideration here is ensuring that the correct macroeconomic variables and their transformations are being picked up in model implementation, to align with the model specifications as written in model documentation. A related consideration is that the functional forms in implementation should correctly mirror that as specified in the approved model equations. I would also add that equally important to the macroeconomic data is the spot portfolio data, that the correct fields corresponding to model development be pulled in, and that the quality of the spot data be checked for anomalies such as invalid or missing values.

What do you see ahead for the future in CECL model implementation and execution?

I believe that the future lies in technology that will allow for more seamless transition of models from the development to the implementation and execution environments, with minimal manual interventions and other patches. Another advance will be in the arena of testing technologies that will allow non-technical stakeholders to glean insights without the intervention of technical or IT staff – for example, in real time being able to test how CECL reserves change with alternative model specifications or segmentations.