Check your toolbox before chasing new (and expensive) modeling systems
The coronavirus pandemic has triggered many banks to reevaluate and upgrade their models, resulting in higher costs to purchase new modeling systems from vendors.
Often overlooked in this process are the various models and analytical tools already within the bank that aren’t being used to their full potential — a savings banks could really use in times of crisis.
More banks should take a top-down approach by looking at all the models across business lines holistically, and how those models can be reused to meet multiple business needs. This approach allows banks to get the most out of their existing intellectual property, and to identify processes that could be improved though models (particularly those that worked in silos) and whether there is a need to develop a new model.
Many banks are already adopting a top-down approach to identify key bank operations and important outputs, such as regulatory reporting. A group called the Bank Industry Architecture Network has identified more than 300 business capabilities, including 12 families of models, that enable a bank’s operations such as underwriting and payments.
The model risk groups validate whether the right models are being used and if it is free from errors. Now, model risk groups are also identifying where a model should be used to enhance a process, in addition to traditional model testing.
Furthermore, when models are seen more as business capabilities, it can be wrapped in application programming interfaces and become part of a broader automated processes. Once a model is implemented within an API framework, it allows the model to be easily reused in other processes, known in the tech world as extensibility.
This approach has pitfalls, but also creates efficiencies for model risk management.
For example, when oil prices were falling amid the global pandemic, many banks used the natural language processing models to analyze the covenants of loan contracts with energy companies. Rather than start from scratch or purchase a new system/model, banks first looked at their existing models and repurposed those to analyze the risk in these contracts.
During this process, many banks’ model risk groups quickly identified issues with the natural language processing models and the model developers addressed the issues in order to swiftly deploy a new risk management process for the emerging crisis. This top-down approach made it easier for lines of business to know about existing models and discuss how to use it for new purposes.
However, model risk groups must maintain independence and not become too prescriptive to the lines of business on which models they must choose. This can be avoided by model risk groups identifying gaps in each model capability, such as natural language processing, without stipulating a specific type of natural language model that must be used.
Also, banks need to ensure that each new use of a model is appropriately evaluated by the model risk group prior to use. And this approach needs to be combined with more traditional ways of identifying existing models to ensure that none are left out of the inventory.
Jacob Kosoff is a senior vice president and head of model risk management and validation at Regions Bank. Aaron Bridgers is a senior vice president and the head of risk testing optimization at Regions Bank. Henry Lee, Ph.D., is a senior risk and financial intelligence consultant at the SAS Institute.
The opinions expressed in the article are statements of the authors’ opinion, are intended only for informational purposes, and are not formal opinions of, nor binding on Regions Bank, its parent company, Regions Financial Corp. and their subsidiaries, and any representation to the contrary is expressly disclaimed.