Santa Fe, NM 87501
Model Risk Management, 23 Jun 2015
We're glad to share with you the executive summary of our article to appear in RMA journal 2015. The topic is best practices in model risk management, and you can access the entire article here. Enjoy!
The article defines the role of the key players in MRM: model developers, model validation, internal audit, and the Model Risk Committee. We explore the scale of each role and the effective way to engage each group within the CCAR framework, which is unique due to the deadlines imposed on an otherwise on-going model development process. We discuss how model validation and internal audit can make a significant impact on the models through iterative touch points and well-defined feedback checklists (provided in the article).
The article introduces the various classes of models required by the regulators: champion models, challenger models, confirming models and validation confirming models. Further, we discuss how model validation and internal audit should grade the models. We argue that a letter-grade system is better than a pass/fail system in today's environment. We also explore the boundaries between validation and audit in terms of grading, per the regulators' requirements. We believe that with all the efforts from the banks, the regulators also need to bring more clarity in their guidelines. Although we acknowledge that today's regulators are much more sophisticated, we do believe that they are also searching for the best answers, along with the banks. Therefore, a more open discussion from both sides would hugely benefit the industry. We have gathered the following key insights in our conclusion:
1. Models need grades rather than pass/fail.
2. Pre-determine the failure criteria.
3. Don't wait until the end. Validate and audit as you go. Give feedback early enough for remediation. Make sure that the models address the regulators' MRIA's (Matters Requiring Immediate Attention) and MRA's (Matters Requiring Attention).
4. Hire the right staff. Banks need quants for both validation and audit.