Effectively managing expanded model inventories
Konstantina Armata, Senior Modelling expert, former Group Head of Model Risk Management, Barclays
Below is an insight into what can be expected from Konstantina’s session at Risk EMEA 2023
The views and opinions expressed in this article are those of the thought leader as an individual, and are not attributed to CeFPro or any particular organization.
What are the main challenges of validating new models?
When the new model to be validated is of the same type as other models previously validated by the team (e.g. a new valuation or credit risk model), then the main challenge is the additional demand on resourcing and the need to reshuffle validation priorities. However, if the new model is of a different model-type, then the challenge is significantly higher. Not only does the team need to reprioritize demand, it also needs to build the capability, framework and skillset required to validate the new type of model. Consider for example Climate Risk Stress Test models which aim to model climate phenomena in conjunction with different socioeconomic, technology and policy trends. These models are very different in nature than other Stress Test models; they are characterized by lack of data and know-how and the usual MRM framework cannot be followed to its full extent. Consequently, the validation team has to set out a new approach, define what validation for these models means (e.g. how we quality assure the output in view of no reliable benchmarking data), agree new testing standards with developers and build the necessary skillset to be able to challenge the models effectively.
How can financial institutions effectively leverage technology to streamline model inventory processes?
In view of an ever growing Model Inventory (MI), and the various audit-trail requirements across the model life-cycle, the use of technology to manage the MI is an imperative. A global firm with a mixture of Retails and IB activities has thousands of models that need to be registered in the MI. Each of these models has tens of attributes that need to be completed and continuously updated and tracked in the MI (e.g. Limitations, validation findings, usage etc). Moreover, a regular MI attestation process involving all Model Owners & Users across the Bank is necessary to keep the MI comprehensive and up-to date. As such, a robust work-flow technology solution is absolutely necessary. The tool needs to be able to reflect complex model interconnectedness relations (feeder, parent, component ), specific usages (e.g. BAU versus stress test), accountable parties, model tiering etc as well as allow workflow capability between the different stakeholders who support the model’s life-cycle.
Why is it important for firms to assess the maturity of their model inventory (MI)?
Because the Model Risk Management landscape is constantly evolving.
Firstly, regarding the model population, new models and model types keep coming under governance, and also different usages of models keep appearing and therefore, it is important to be able to capture those in a timely manner.
Second, as a result of model monitoring activities, revalidation etc., new issues with the models get discovered, prior issues get fixed, the tiering of the model changes etc., and so again, it is important to keep assessing the MI to keep the information current.
Finally the MI and the information contained therein, informs a number of model life-cycle activities such as revalidation frequency, escalation of overdue remediation, model risk reports to the Board etc. Therefore, outdated information will not only cause inefficiencies but it may also be misguiding in terms of the level of model risk at any given point in time.
How can model inventory be used to support different functions?
The workflow tool I referred to previously is designed to support every function involved in the model life-cycle. It starts with a new model submission request by the model owner complemented by a model development documentation submitted by the model developer. The workflow then moves to the relevant validation team member and the validation process gets allocated and commences. There will be different workflow updates during the validation process with the workflow moving between model developer, owner and validation. The same holds true throughout the model’s life cycle post validation as the MI will keep getting updated with information that emerges as a result of model changes, model performance monitoring and other activities and so for as long as a model exists in the MI, all different parties will keep engaging with the MI tool.
What are the benefits of moving from an inventory to a model risk management system?
An inventory is a snap-shot, lifeless tool that worked in the past when model validation was a one-off and, perhaps the only model governance activity required. However, nowadays model risk management (as required by SR11-7 or CP6-22) is a lot broader than just model validation and it requires ongoing activities across the life-cycle involving various touchpoints between all relevant stakeholders. Therefore, for the reasons mentioned above, a model risk management workflow tool is necessary.