• Anti-money Laundering
  • Report
  • Insurance
  • Artificial Intelligence
  • MRM function
  • Global Research And Analytics
May 08, 2020

Panel discussion on Implementing Model Risk Management practices at insurers

Key findings


CRISIL GR&A hosted a roundtable with top model risk management (MRM) practice leaders in the insurance industry in New York. We welcomed more than 30 attendees. A panel was formed by Nikolai Kukharkin (Head of Model Risk at TIAA), Phil Elam (Head of Model Risk at Prudential Financial), Mark Kust (Model Risk Expert at VCP), Rodanthy Tzani (Head of Model Risk at New York life) and Eric Tam (Managing Director for Model Risk at AIG). The panel was moderated by Alberto Ramirez, Practice Leader (Insurance), CRISIL.


Adoption of MRM by insurers


While the US banking industry embraced MRM programs since the early 2000s and more rigorously from 2011 (after the issuance of SR 11-7, a Supervisory Letter (SR) guidance on MRM from the US Federal Reserve and the Office of the Comptroller of the Currency [OCC]), the insurance industry started adopting MRM programs only around 2016. Since then models have been classified based on model components, with many pieces nested within the models.


…the Fed expanded the MRM focus to all models


The 2008 financial crisis increased the focus on risk models – from pricing to anti-money laundering (AML) risk models – which account to 30% of models. Initially insurers were focused on actuarial models, but the Fed expanded the MRM focus to all models.


Insurers define models or model components in different ways. Depending on the definition, they can get an inventory of 300 to 3,000 models. Model components reduce model complexity, but might increase the effort to establish inherent risk and dependencies while performing model validation. A thorough review is necessary to determine which models should be considered and which ones should be excluded from the model inventory, based on the nature of their use, such as end-user computing tools (EUCTs). It is also important to determine the linkages or interdependencies of models.


Even though insurers have their own model inventories, it is important to re-evaluate the completeness of their inventories, as a few models might have been left out unintentionally. Model inventories are never complete and there are no guidelines for insurers on this currently.


Over the past three years, industry practices regarding models have changed. There have been adjustments in the validation cycle from one year to the next, and the number of models have reduced. Moreover, many model changes are occurring due to regulations (principle-based reserving), and the trend is likely to continue. As of now, US insurers have 60% actuarial models, 20% financial models, and the rest risk models from corporate.