Reviewing the evolution of model risk management and managing innovation with management and oversight

By Sophie Bottazzi, Senior Research Executive, CeFPro

Innovation and advances to model risk management have evolved rapidly over the last few years with new techniques and opportunities available. As management continues to evolve, challenges with compliance, competition and validation arise for financial institutions. As the industry continues to evolve, it is a race fro innovation to provide the optimum customer experience and limit human manual processes. Model risk has traditionally been very a very human labor intensive exercise with manual input and validation, can the industry move towards reliance on models and ensure trust in outputs?

To explore these themes further The Center for Financial Professionals will host the 2nd Annual Model Risk Management USA Congress, taking place October 7-8 in New York City. The two-day congress will review the latest challenges, opportunities and innovation in Model Risk Management, including a one day forum focusing entirely on managing and validating machine learning models led by industry experts from Wells Fargo, Dr. Agus Sudjianto and his team.

As with many aspects of financial risk, compliance and keeping on top of regulatory change and expectations is a continual focus. As model risk continues to evolve, regulatory bodies increased focus and attention on ensuring safety, soundness and accuracy of models and outputs. A big focus within financial institutions is on leveraging technology uses and increasing efficiency, however are long standing regulations applicable in this new era and how applicable are the one size fits all mandates? Many institutions are turning their focus to implementing AI and Machine Learning capabilities and determining the most efficient ways to validate and control these models and the risk associated with any algorithm models. Institutions remain in a high pressured environment to always ensure cost efficiency though are institutions willing to make the investment needed to then see the potential of long term efficiency? Institutions must look to develop an enhanced governance and oversight framework to ensure effective reviews and validation of algorithm models. An ongoing concern relating to AI and machine learning models comes that of bias in the output and the question of how to remove this. The reputational risks associated with any form of bias in these algorithms could be unquantifiable in nature. US Lawmakers are looking to address AI and machine learning bias with the Algorithmic Accountability Act which looks to increase transparency and interpretability of AI and machine learning models. This Act could assist regulatory bodies in financial services to oversee the use of these emerging models and ensure their accuracy and soundness to the financial system.

A second area for discussion across the industry and at the 2nd Annual Model Risk Management USA Congress was that of qualitative models and understanding the what and why of these models. Increasingly qualitative models are being brought into the scope of model risk teams who are seeing an expended definition of a model from regulators. Model risk experts are grappling with managing assumptions under qualitative modeling and standard testing requirements to justify assumptions. The industry looks to the regulation for guidance as to development and validation techniques for a widely unfamiliar model. With qualitative models increasing in reliance and importance within institutions it becomes vital for a complete inventory of them to ensure effective review, validation and monitoring. Many are looking to understand the treatment of qualitative models and the scrutiny required with less quantitative input and output.

Finally, a key element that was repeated throughout the research is the transition to CECL and managing this to get to go live day 1. As SEC filers move ever closer to implementation, regulators and the industry look to draw leanings and lessons from the implementation process. Aligned with the aforementioned topic, qualitative overlays remain an unfamiliar territory and look to have an increased impact on reserves, therefore remaining ever vital to the financial outlook of an institution. Many institutions are leveraging CCAR models as part of their implementation of CECL, but how to they align the two and ensure that qualitative overlays and assumptions can be made for CECL and remain independent of CCAR processes?

Overall there has been a high focus on regulatory expectations with innovation in mind and keeping ahead. The industry needs to develop the model risk management practices tailored to individual model requirements, as they are implemented into the future state of model risk management. With the evolution of model definition, their uses and management; institutions must stay vigilant over innovation and product offerings.

The 2nd Annual Model Risk Management Congress looks to address the above challenges and much more, providing a platform for best practice and idea sharing to enhance collaboration and communication. The findings of this research will be illustrated on October 7-8, 2019 at CeFPro’s 2nd Annual Model Risk Management Congress in New York City. We invite you to join your peers for two days to discuss upcoming regulation, AI & Machine Learning, Qualitative Models, Validation, CECL The future of model risk management and many more. For further information, please get in touch with a member of the team on +1 888 677 7007.