Ever since Harry Markowitz introduced modern portfolio theory, or mean-variance analysis, in 1952, banking has relied on advanced mathematics to automate and improve decision-making. While risk-based portfolio management is the most famous example of mathematical models penetrating banking, today they are deployed across the entire bank to help optimize everything from asset/liability management (ALM) to anti-money laundering (AML) activities.
While both the past and future of analytical quantitative models in financial services are extremely rich in scope, there are key trends shaping the future of model risk management that cannot be ignored. In a session entitled “The Shape of Model Risk Management to Come” at the 2020 RMA Annual Risk Management Virtual Conference, I focused on the future of how models will be utilized and managed at banks and other institutions as theoretical innovations clash and combine with technological innovations. Here are the key topics I covered in the 25-minute presentation:
Compliance and Modeling
In recent times, compliance has ballooned into a multi-regulatory, cross-disciplinary exercise spanning many jurisdictions within an institution. Today, institutions must prioritize model development and validation on a risk-adjusted basis to get the most out of their limited resources while meeting dynamic, quickly evolving compliance requirements and expectations. At the same time, institutions are exploring new tools like natural language processing to digest regulations and model outputs to optimize the institutional response.
AML/Fraud Detection Modeling
As opportunities to perpetrate money laundering and other financial crimes expand, banks must develop their own tools and strategies to keep pace. In the presentation, I took stock of a few of these model-based tools being used to identify fraud, including big data and statistical analysis, artificial intelligence (AI)/machine learning (ML) solutions, and social network analysis.
Impact of LIBOR Transition
Another “greatest hits” topic in modern model risk management is the transition from LIBOR to a new interest reference rate. I highlighted how there will likely not be just one alternative risk-free rate prevailing in the U.S., surveyed how stress testing and CCAR will be impacted by the transition, and unpacked related topics like non-cleared derivatives, rate basis risk, and the volatility surface.
Presenting on the future of model risk management at the 2020 RMA Annual Risk Management Virtual Conference
AI/ML models may be relatively new to the world of finance, but the rules for developing and validating them are mostly the same as the old rules corresponding to more traditional models. At the virtual conference, I detailed the importance of AI/ML model selection, accuracy, and stability and explored how the underlying cause and effect are key to conceptual soundness.
I also touched on how many emerging technologies, such as chatbots, biometric recognition solutions, AI/ML-assisted marketing tools, are actually models and should be treated as such, meaning that they deserve the same scrutiny of methodology and reliability.
Banking in the Digital Age
Finally, I discussed banking in the digital age and how enabling a best-in-class customer experience isn’t just a technology or digital experience design challenge. It also relies on quality advanced mathematical analysis to enable real-time know your customer (KYC) proceedings, transaction analysis, and reporting. It will require more sophisticated models, algorithms, and related technologies to do this well and successfully transition toward the shape of model risk management to come.
As the Managing Director of the RMA Model Validation Consortium, Kevin is passionate about providing high-quality model validation services at a competitive price point for RMA member banks. Kevin holds a Ph.D. in math from UCLA and was a leader in risk management and model validation for Wells Fargo Bank.