Regulatory changes under Model Risk Management: When should insurers start preparing?

Our previous blog in July 2024 summarised the Implications on Insurers of Model Risk Management principles for banks (SS1/23). We expect the PRA to formally extend the MRM principles to insurers during 2025. Therefore, it is imperative that insurers look to understand the implications of these principles on their existing model risk frameworks and consider whether there are any tech-based solutions available in the market that could be used to adhere to the principles efficiently.

Our conversations with insurers in 2024 around MRM have provided insight into the different challenges they may face across each of the principles outlined in our previous blog post. Outside of IFRS 17 and Solvency II reporting processes, insurance companies vary significantly in their preparedness to define and identify models across their entire operations, with many having no agreed definition even across different departments within the same organisation. A summary of the key themes, by principle, is included below:

• Principle 1 – Model Identification and Model Risk Classification: All functions of an insurance company will need to consider their definition of models and Deterministic Quantitative methods 'DQMs', including functions which have historically never come under the scope of model risk. Using clear definitions, clear questionnaires and decision trees as part of the model identification process has made the process significantly easier.
• Principle 2 – Governance: Insurers are finding initial resistance and friction in ensuring model risk management is understood and sponsored at all senior levels. Senior management must invest in learning and upskilling sessions to ensure compliance.
• Principle 3 – Model Development, Implementation and Use: Model development is an area where insurers typically excel. However, many remain uncertain about the required documentation and testing levels for different model tiers. Still, there are certain insurers in the market who need to enhance their model development process.
• Principle 4 – Independent Model Validation: There will be significant resource challenges faced by insurers to meet the increased level and depth of model validation required under SS1/23. Best practices for validating AI in models remain undefined, creating an additional challenge for insurers.
• Principle 5 – Model Risk Mitigants: Insurers lack consensus on defining model risk policies for 'Out-of-model adjustments' (OOMAs), including criteria for their appropriateness. Similar little consensus surrounds how companies should determine the length of time for which an OOMA must be brought into larger valuation/pricing models.

Based on feedback received from the banking clients around their early journey on MRM implementation and the associated tools available, we developed our own in-house solution, Model Control Centre (MCC), as a useful solution to help empower clients and enable seamless MRM compliance. Based on the key themes emerging from insurers, we believe that insurers might find MCC or a similar solution useful to leverage to meet the PRA guidelines. We deploy MCC on all our model and model related validation projects which we find enables a seamless planning, tracking and reporting on the project for clients and regulators.

MCC is a cloud-based platform which is equipped to assist insurers in simplifying and strengthening model risk management by integrating model inventory, development & validation activities and reporting of key findings in an automated manner. It can manage all your organisation's models and is scalable to accommodate annual growth making it an ideal solution for expanding insurance businesses. The platform offers a rapid deployment with customized configurations and training of staff, to meet specific business needs.

The main functionalities of MCC are:

• Model Inventory: ‘All in one place’ repository for tracking and managing all models with respective risk ratings and other associated details.?
• Validation Components: Allows scheduling of validation and periodic re-validation procedures for each model linked to their risk rating, with designated validators and approvers.?
• Findings: Logs validation findings with materiality ratings and remediation status, serving as a single source for validation results.
• Reporting: Real-time dashboards for model inventory and status of validation activity assigned to each validator, reviewer, or approver. Also, generates an automated executive summary of findings.

The essential features of MCC enables users to maintain a comprehensive model inventory encompassing all of the organisation’s models from development stage to the ones decommissioned. Further, scheduling of tasks with pre-configured procedures and documentation of findings ensures consistency in both the development and validation processes across the organisation, while also highlighting critical issues in a structured way. This level of automation by MCC, significantly boosting efficiency and frees up resources to focus on high-value tasks.

Additional MRM Tools and Accelerators

Through our experience across MRM framework implementation in the banking sector, we have developed the following additional tools in response to what banks have found most useful in their journey to becoming SS1/23 compliant. These tools are industry agnostic and can also be used to help insurers with their MRM framework:

• Gold standard MRM frameworks, policy, development and validation standards: These accelerators include comprehensive model development and validation of standards for a variety of models.
• Board and business workshop and education material: Accelerators provide off the shelf workshop materials to run sessions with the board and all business units to upskill on the requirements of SS1/23.
• SS1/23 self assessment questionnaire: A comprehensive questionnaire for insurers to use as an initial self assessment against MRM policies and SS1/23 guidelines. This tool also showcases the level of MRM effectiveness across models.
• Automated Deterministic Quantitative Method (DQM) identifier: PwC’s EUCA (End User Computer Analytics) technology is a solution that allows insurers to identify, discover and visualise DQMs across an organisation in a systematic fashion, allowing for comprehensive model inventory analysis.
• Model tiering assessment: This tool allows insurers to define and assess different model risk criteria at a granular level and derive final model tiers based upon these criteria.
• Model risk appetite metrics and reporting metrics templates: These tools provide ready made templates for quantitative and qualitative measures that help to articulate and quantify the level of model risk.
• Model agnostic Post-Model Adjustment (PMA) framework: This tool allows insurers to embed a consistent MRM framework for PMAs covering adjustment lifecycle, governance and approvals.

As we wait for PRA to extend Model Risk Management principles to insurers, it is crucial for insurance companies to proactively prepare for these regulatory changes. Early preparation and strategic investment in MRM solutions will position insurers to not only comply with upcoming regulations but also enhance operational efficiency and risk management capabilities.